[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/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 = GFS2_I(jd->jd_inode);
	struct gfs2_glock *gl = ip->i_gl;
	int new = 0;
	uint64_t dblock;
	uint32_t 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;
	}

	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(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 *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(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 *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_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 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 *head)
{
	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	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_inode, 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_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;
}

/**
 * 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 = GFS2_I(jd->jd_inode);
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	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	/*
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	*/
	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	{
feaa7bba	35	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
5c676f6d	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
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
	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 \|\|
59a1cc6b	156	lh.lh_blkno != blk \|\| lh.lh_hash != hash)
b3b94faa DT	157	return 1;
	158
	159	*head = lh;
	160
	161	return 0;
	162	}
	163
	164	/**
	165	* find_good_lh - find a good log header
	166	* @jd: the journal
	167	* @blk: the segment to start searching from
	168	* @lh: the log header to fill in
	169	* @forward: if true search forward in the log, else search backward
	170	*
	171	* Call get_log_header() to get a log header for a segment, but if the
	172	* segment is bad, either scan forward or backward until we find a good one.
	173	*
	174	* Returns: errno
	175	*/
	176
	177	static int find_good_lh(struct gfs2_jdesc jd, unsigned int blk,
	178	struct gfs2_log_header *head)
	179	{
	180	unsigned int orig_blk = *blk;
	181	int error;
	182
	183	for (;;) {
	184	error = get_log_header(jd, *blk, head);
	185	if (error <= 0)
	186	return error;
	187
	188	if (++*blk == jd->jd_blocks)
	189	*blk = 0;
	190
	191	if (*blk == orig_blk) {
feaa7bba	192	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	193	return -EIO;
	194	}
	195	}
	196	}
	197
	198	/**
	199	* jhead_scan - make sure we've found the head of the log
	200	* @jd: the journal
	201	* @head: this is filled in with the log descriptor of the head
	202	*
	203	* At this point, seg and lh should be either the head of the log or just
	204	* before. Scan forward until we find the head.
	205	*
	206	* Returns: errno
	207	*/
	208
	209	static int jhead_scan(struct gfs2_jdesc jd, struct gfs2_log_header head)
	210	{
	211	unsigned int blk = head->lh_blkno;
	212	struct gfs2_log_header lh;
	213	int error;
	214
	215	for (;;) {
	216	if (++blk == jd->jd_blocks)
	217	blk = 0;
	218
	219	error = get_log_header(jd, blk, &lh);
	220	if (error < 0)
	221	return error;
	222	if (error == 1)
	223	continue;
	224
	225	if (lh.lh_sequence == head->lh_sequence) {
feaa7bba	226	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	227	return -EIO;
	228	}
	229	if (lh.lh_sequence < head->lh_sequence)
	230	break;
	231
	232	*head = lh;
	233	}
	234
	235	return 0;
	236	}
	237
	238	/**
	239	* gfs2_find_jhead - find the head of a log
	240	* @jd: the journal
	241	* @head: the log descriptor for the head of the log is returned here
	242	*
	243	* Do a binary search of a journal and find the valid log entry with the
	244	* highest sequence number. (i.e. the log head)
	245	*
	246	* Returns: errno
	247	*/
	248
	249	int gfs2_find_jhead(struct gfs2_jdesc jd, struct gfs2_log_header head)
	250	{
	251	struct gfs2_log_header lh_1, lh_m;
	252	uint32_t blk_1, blk_2, blk_m;
	253	int error;
	254
	255	blk_1 = 0;
	256	blk_2 = jd->jd_blocks - 1;
	257
	258	for (;;) {
	259	blk_m = (blk_1 + blk_2) / 2;
	260
	261	error = find_good_lh(jd, &blk_1, &lh_1);
	262	if (error)
	263	return error;
	264
	265	error = find_good_lh(jd, &blk_m, &lh_m);
	266	if (error)
	267	return error;
	268
	269	if (blk_1 == blk_m \|\| blk_m == blk_2)
	270	break;
	271
	272	if (lh_1.lh_sequence <= lh_m.lh_sequence)
	273	blk_1 = blk_m;
	274	else
	275	blk_2 = blk_m;
	276	}
	277
	278	error = jhead_scan(jd, &lh_1);
	279	if (error)
	280	return error;
	281
	282	*head = lh_1;
	283
	284	return error;
	285	}
	286
	287	/**
	288	* foreach_descriptor - go through the active part of the log
	289	* @jd: the journal
	290	* @start: the first log header in the active region
291	* @end: the last log header (don't process the contents of this entry))
292	*
293	* Call a given function once for every log descriptor in the active
294	* portion of the log.
295	*
296	* Returns: errno
297	*/
298
299	static int foreach_descriptor(struct gfs2_jdesc *jd, unsigned int start,
300	unsigned int end, int pass)
301	{
feaa7bba	302	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	303	struct buffer_head *bh;
	304	struct gfs2_log_descriptor *ld;
	305	int error = 0;
	306	u32 length;
	307	__be64 *ptr;
	308	unsigned int offset = sizeof(struct gfs2_log_descriptor);
	309	offset += (sizeof(__be64)-1);
	310	offset &= ~(sizeof(__be64)-1);
	311
	312	while (start != end) {
	313	error = gfs2_replay_read_block(jd, start, &bh);
	314	if (error)
	315	return error;
	316	if (gfs2_meta_check(sdp, bh)) {
	317	brelse(bh);
	318	return -EIO;
	319	}
	320	ld = (struct gfs2_log_descriptor *)bh->b_data;
	321	length = be32_to_cpu(ld->ld_length);
	322
e3167ded	323	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
b3b94faa DT	324	struct gfs2_log_header lh;
	325	error = get_log_header(jd, start, &lh);
	326	if (!error) {
	327	gfs2_replay_incr_blk(sdp, &start);
88721877	328	brelse(bh);
b3b94faa DT	329	continue;
	330	}
	331	if (error == 1) {
feaa7bba	332	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	333	error = -EIO;
	334	}
	335	brelse(bh);
	336	return error;
	337	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	338	brelse(bh);
	339	return -EIO;
	340	}
	341	ptr = (__be64 *)(bh->b_data + offset);
	342	error = lops_scan_elements(jd, start, ld, ptr, pass);
	343	if (error) {
	344	brelse(bh);
	345	return error;
	346	}
	347
	348	while (length--)
	349	gfs2_replay_incr_blk(sdp, &start);
	350
	351	brelse(bh);
	352	}
	353
	354	return 0;
	355	}
	356
	357	/**
	358	* clean_journal - mark a dirty journal as being clean
	359	* @sdp: the filesystem
	360	* @jd: the journal
	361	* @gl: the journal's glock
	362	* @head: the head journal to start from
	363	*
	364	* Returns: errno
	365	*/
	366
	367	static int clean_journal(struct gfs2_jdesc jd, struct gfs2_log_header head)
	368	{
feaa7bba SW	369	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	370	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	371	unsigned int lblock;
	372	int new = 0;
	373	uint64_t dblock;
	374	struct gfs2_log_header *lh;
	375	uint32_t hash;
	376	struct buffer_head *bh;
	377	int error;
fd88de56 SW	378	int boundary;
fd88de56 SW	379
b3b94faa DT	380	lblock = head->lh_blkno;
b3b94faa DT	381	gfs2_replay_incr_blk(sdp, &lblock);
feaa7bba	382	error = gfs2_block_map(&ip->i_inode, lblock, &new, &dblock, &boundary);
b3b94faa DT	383	if (error)
	384	return error;
	385	if (!dblock) {
	386	gfs2_consist_inode(ip);
	387	return -EIO;
	388	}
	389
	390	bh = sb_getblk(sdp->sd_vfs, dblock);
	391	lock_buffer(bh);
	392	memset(bh->b_data, 0, bh->b_size);
	393	set_buffer_uptodate(bh);
	394	clear_buffer_dirty(bh);
	395	unlock_buffer(bh);
	396
	397	lh = (struct gfs2_log_header *)bh->b_data;
	398	memset(lh, 0, sizeof(struct gfs2_log_header));
	399	lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
88721877	400	lh->lh_header.mh_type = cpu_to_be32(GFS2_METATYPE_LH);
e3167ded	401	lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH);
b3b94faa DT	402	lh->lh_sequence = cpu_to_be64(head->lh_sequence + 1);
	403	lh->lh_flags = cpu_to_be32(GFS2_LOG_HEAD_UNMOUNT);
	404	lh->lh_blkno = cpu_to_be32(lblock);
	405	hash = gfs2_disk_hash((const char *)lh, sizeof(struct gfs2_log_header));
	406	lh->lh_hash = cpu_to_be32(hash);
	407
	408	set_buffer_dirty(bh);
	409	if (sync_dirty_buffer(bh))
	410	gfs2_io_error_bh(sdp, bh);
	411	brelse(bh);
	412
	413	return error;
	414	}
	415
	416	/**
	417	* gfs2_recover_journal - recovery a given journal
	418	* @jd: the struct gfs2_jdesc describing the journal
b3b94faa DT	419	*
	420	* Acquire the journal's lock, check to see if the journal is clean, and
	421	* do recovery if necessary.
	422	*
	423	* Returns: errno
	424	*/
	425
c63e31c2	426	int gfs2_recover_journal(struct gfs2_jdesc *jd)
b3b94faa	427	{
feaa7bba SW	428	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	429	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	430	struct gfs2_log_header head;
	431	struct gfs2_holder j_gh, ji_gh, t_gh;
	432	unsigned long t;
	433	int ro = 0;
	434	unsigned int pass;
	435	int error;
	436
5965b1f4 SW	437	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	438	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
	439	jd->jd_jid);
b3b94faa	440
5965b1f4	441	/* Aquire the journal lock so we can do recovery */
b3b94faa	442
5965b1f4 SW	443	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	444	LM_ST_EXCLUSIVE,
	445	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	446	&j_gh);
	447	switch (error) {
	448	case 0:
	449	break;
	450
	451	case GLR_TRYFAILED:
	452	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	453	error = 0;
	454
	455	default:
	456	goto fail;
	457	};
b3b94faa	458
5965b1f4 SW	459	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
	460	LM_FLAG_NOEXP, &ji_gh);
	461	if (error)
	462	goto fail_gunlock_j;
	463	} else {
	464	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	465	}
b3b94faa DT	466
	467	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	468
	469	error = gfs2_jdesc_check(jd);
	470	if (error)
	471	goto fail_gunlock_ji;
	472
	473	error = gfs2_find_jhead(jd, &head);
	474	if (error)
	475	goto fail_gunlock_ji;
	476
	477	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
	478	fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
	479	jd->jd_jid);
	480
	481	t = jiffies;
	482
	483	/* Acquire a shared hold on the transaction lock */
	484
59a1cc6b	485	error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED,
5965b1f4	486	LM_FLAG_NOEXP \| LM_FLAG_PRIORITY \|
59a1cc6b	487	GL_NOCANCEL \| GL_NOCACHE, &t_gh);
b3b94faa DT	488	if (error)
	489	goto fail_gunlock_ji;
	490
	491	if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
	492	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	493	ro = 1;
	494	} else {
	495	if (sdp->sd_vfs->s_flags & MS_RDONLY)
	496	ro = 1;
	497	}
	498
	499	if (ro) {
	500	fs_warn(sdp, "jid=%u: Can't replay: read-only FS\n",
	501	jd->jd_jid);
	502	error = -EROFS;
	503	goto fail_gunlock_tr;
	504	}
	505
	506	fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);
	507
	508	for (pass = 0; pass < 2; pass++) {
	509	lops_before_scan(jd, &head, pass);
	510	error = foreach_descriptor(jd, head.lh_tail,
	511	head.lh_blkno, pass);
	512	lops_after_scan(jd, error, pass);
	513	if (error)
	514	goto fail_gunlock_tr;
	515	}
	516
	517	error = clean_journal(jd, &head);
	518	if (error)
	519	goto fail_gunlock_tr;
	520
	521	gfs2_glock_dq_uninit(&t_gh);
5c676f6d	522	t = DIV_ROUND_UP(jiffies - t, HZ);
b3b94faa DT	523	fs_info(sdp, "jid=%u: Journal replayed in %lus\n",
	524	jd->jd_jid, t);
	525	}
	526
5965b1f4 SW	527	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	528	gfs2_glock_dq_uninit(&ji_gh);
b3b94faa DT	529
	530	gfs2_lm_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
	531
5965b1f4 SW	532	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	533	gfs2_glock_dq_uninit(&j_gh);
b3b94faa DT	534
b3b94faa DT	535	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
b3b94faa DT	536	return 0;
b3b94faa DT	537
5965b1f4	538	fail_gunlock_tr:
b3b94faa	539	gfs2_glock_dq_uninit(&t_gh);
5965b1f4 SW	540	fail_gunlock_ji:
	541	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	542	gfs2_glock_dq_uninit(&ji_gh);
	543	fail_gunlock_j:
	544	gfs2_glock_dq_uninit(&j_gh);
	545	}
b3b94faa DT	546
	547	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
	548
5965b1f4	549	fail:
b3b94faa	550	gfs2_lm_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
b3b94faa DT	551	return error;
	552	}
	553
	554	/**
	555	* gfs2_check_journals - Recover any dirty journals
	556	* @sdp: the filesystem
	557	*
	558	*/
	559
	560	void gfs2_check_journals(struct gfs2_sbd *sdp)
	561	{
	562	struct gfs2_jdesc *jd;
	563
	564	for (;;) {
	565	jd = gfs2_jdesc_find_dirty(sdp);
	566	if (!jd)
	567	break;
	568
	569	if (jd != sdp->sd_jdesc)
c63e31c2	570	gfs2_recover_journal(jd);
b3b94faa DT	571	}
	572	}
	573