[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/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 <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 = {
	.owner	 = THIS_MODULE,
	.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
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>
fe64d517	17	#include <linux/slow-work.h>
b3b94faa DT	18
b3b94faa DT	19	#include "gfs2.h"
5c676f6d	20	#include "incore.h"
b3b94faa DT	21	#include "bmap.h"
	22	#include "glock.h"
	23	#include "glops.h"
b3b94faa DT	24	#include "lops.h"
	25	#include "meta_io.h"
	26	#include "recovery.h"
	27	#include "super.h"
5c676f6d	28	#include "util.h"
71b86f56	29	#include "dir.h"
b3b94faa DT	30
	31	int gfs2_replay_read_block(struct gfs2_jdesc *jd, unsigned int blk,
	32	struct buffer_head **bh)
	33	{
feaa7bba	34	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
5c676f6d	35	struct gfs2_glock *gl = ip->i_gl;
b3b94faa	36	int new = 0;
cd915493 SW	37	u64 dblock;
cd915493 SW	38	u32 extlen;
b3b94faa DT	39	int error;
b3b94faa DT	40
feaa7bba	41	error = gfs2_extent_map(&ip->i_inode, blk, &new, &dblock, &extlen);
b3b94faa DT	42	if (error)
	43	return error;
	44	if (!dblock) {
5c676f6d	45	gfs2_consist_inode(ip);
b3b94faa DT	46	return -EIO;
	47	}
	48
7276b3b0	49	*bh = gfs2_meta_ra(gl, dblock, extlen);
b3b94faa DT	50
	51	return error;
	52	}
	53
cd915493	54	int gfs2_revoke_add(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
b3b94faa DT	55	{
	56	struct list_head *head = &sdp->sd_revoke_list;
	57	struct gfs2_revoke_replay *rr;
	58	int found = 0;
	59
	60	list_for_each_entry(rr, head, rr_list) {
	61	if (rr->rr_blkno == blkno) {
	62	found = 1;
	63	break;
	64	}
	65	}
	66
	67	if (found) {
	68	rr->rr_where = where;
	69	return 0;
	70	}
	71
16c5f06f	72	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
b3b94faa DT	73	if (!rr)
	74	return -ENOMEM;
	75
	76	rr->rr_blkno = blkno;
	77	rr->rr_where = where;
	78	list_add(&rr->rr_list, head);
	79
	80	return 1;
	81	}
	82
cd915493	83	int gfs2_revoke_check(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
b3b94faa DT	84	{
	85	struct gfs2_revoke_replay *rr;
	86	int wrap, a, b, revoke;
	87	int found = 0;
	88
	89	list_for_each_entry(rr, &sdp->sd_revoke_list, rr_list) {
	90	if (rr->rr_blkno == blkno) {
	91	found = 1;
	92	break;
	93	}
	94	}
	95
	96	if (!found)
	97	return 0;
	98
	99	wrap = (rr->rr_where < sdp->sd_replay_tail);
	100	a = (sdp->sd_replay_tail < where);
	101	b = (where < rr->rr_where);
	102	revoke = (wrap) ? (a \|\| b) : (a && b);
	103
	104	return revoke;
	105	}
	106
	107	void gfs2_revoke_clean(struct gfs2_sbd *sdp)
	108	{
	109	struct list_head *head = &sdp->sd_revoke_list;
	110	struct gfs2_revoke_replay *rr;
	111
	112	while (!list_empty(head)) {
	113	rr = list_entry(head->next, struct gfs2_revoke_replay, rr_list);
	114	list_del(&rr->rr_list);
	115	kfree(rr);
	116	}
	117	}
	118
bb8d8a6f SW	119	static int gfs2_log_header_in(struct gfs2_log_header_host lh, const void buf)
	120	{
	121	const struct gfs2_log_header *str = buf;
	122
	123	if (str->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) \|\|
	124	str->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH))
	125	return 1;
	126
	127	lh->lh_sequence = be64_to_cpu(str->lh_sequence);
	128	lh->lh_flags = be32_to_cpu(str->lh_flags);
	129	lh->lh_tail = be32_to_cpu(str->lh_tail);
	130	lh->lh_blkno = be32_to_cpu(str->lh_blkno);
	131	lh->lh_hash = be32_to_cpu(str->lh_hash);
	132	return 0;
	133	}
	134
b3b94faa DT	135	/**
	136	* get_log_header - read the log header for a given segment
	137	* @jd: the journal
	138	* @blk: the block to look at
	139	* @lh: the log header to return
	140	*
	141	* Read the log header for a given segement in a given journal. Do a few
	142	* sanity checks on it.
	143	*
	144	* Returns: 0 on success,
	145	* 1 if the header was invalid or incomplete,
	146	* errno on error
	147	*/
	148
	149	static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
55167622	150	struct gfs2_log_header_host *head)
b3b94faa DT	151	{
b3b94faa DT	152	struct buffer_head *bh;
f5a8cd02	153	struct gfs2_log_header_host uninitialized_var(lh);
887bc5d0	154	const u32 nothing = 0;
cd915493	155	u32 hash;
b3b94faa DT	156	int error;
	157
	158	error = gfs2_replay_read_block(jd, blk, &bh);
	159	if (error)
	160	return error;
	161
2a2c9824 SW	162	hash = crc32_le((u32)~0, bh->b_data, sizeof(struct gfs2_log_header) -
	163	sizeof(u32));
	164	hash = crc32_le(hash, (unsigned char const *)&nothing, sizeof(nothing));
	165	hash ^= (u32)~0;
bb8d8a6f	166	error = gfs2_log_header_in(&lh, bh->b_data);
b3b94faa DT	167	brelse(bh);
b3b94faa DT	168
bb8d8a6f	169	if (error \|\| lh.lh_blkno != blk \|\| lh.lh_hash != hash)
b3b94faa DT	170	return 1;
	171
	172	*head = lh;
	173
	174	return 0;
	175	}
	176
	177	/**
	178	* find_good_lh - find a good log header
	179	* @jd: the journal
	180	* @blk: the segment to start searching from
	181	* @lh: the log header to fill in
	182	* @forward: if true search forward in the log, else search backward
	183	*
	184	* Call get_log_header() to get a log header for a segment, but if the
	185	* segment is bad, either scan forward or backward until we find a good one.
	186	*
	187	* Returns: errno
	188	*/
	189
	190	static int find_good_lh(struct gfs2_jdesc jd, unsigned int blk,
55167622	191	struct gfs2_log_header_host *head)
b3b94faa DT	192	{
	193	unsigned int orig_blk = *blk;
	194	int error;
	195
	196	for (;;) {
	197	error = get_log_header(jd, *blk, head);
	198	if (error <= 0)
	199	return error;
	200
	201	if (++*blk == jd->jd_blocks)
	202	*blk = 0;
	203
	204	if (*blk == orig_blk) {
feaa7bba	205	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	206	return -EIO;
	207	}
	208	}
	209	}
	210
	211	/**
	212	* jhead_scan - make sure we've found the head of the log
	213	* @jd: the journal
	214	* @head: this is filled in with the log descriptor of the head
	215	*
	216	* At this point, seg and lh should be either the head of the log or just
	217	* before. Scan forward until we find the head.
	218	*
	219	* Returns: errno
	220	*/
	221
55167622	222	static int jhead_scan(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa DT	223	{
b3b94faa DT	224	unsigned int blk = head->lh_blkno;
55167622	225	struct gfs2_log_header_host lh;
b3b94faa DT	226	int error;
	227
	228	for (;;) {
	229	if (++blk == jd->jd_blocks)
	230	blk = 0;
	231
	232	error = get_log_header(jd, blk, &lh);
	233	if (error < 0)
	234	return error;
	235	if (error == 1)
	236	continue;
	237
	238	if (lh.lh_sequence == head->lh_sequence) {
feaa7bba	239	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	240	return -EIO;
	241	}
	242	if (lh.lh_sequence < head->lh_sequence)
	243	break;
	244
	245	*head = lh;
	246	}
	247
	248	return 0;
	249	}
	250
	251	/**
	252	* gfs2_find_jhead - find the head of a log
	253	* @jd: the journal
	254	* @head: the log descriptor for the head of the log is returned here
	255	*
	256	* Do a binary search of a journal and find the valid log entry with the
	257	* highest sequence number. (i.e. the log head)
	258	*
	259	* Returns: errno
	260	*/
	261
55167622	262	int gfs2_find_jhead(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa	263	{
55167622	264	struct gfs2_log_header_host lh_1, lh_m;
cd915493	265	u32 blk_1, blk_2, blk_m;
b3b94faa DT	266	int error;
	267
	268	blk_1 = 0;
	269	blk_2 = jd->jd_blocks - 1;
	270
	271	for (;;) {
	272	blk_m = (blk_1 + blk_2) / 2;
	273
	274	error = find_good_lh(jd, &blk_1, &lh_1);
	275	if (error)
	276	return error;
	277
	278	error = find_good_lh(jd, &blk_m, &lh_m);
	279	if (error)
	280	return error;
	281
	282	if (blk_1 == blk_m \|\| blk_m == blk_2)
	283	break;
	284
	285	if (lh_1.lh_sequence <= lh_m.lh_sequence)
	286	blk_1 = blk_m;
	287	else
	288	blk_2 = blk_m;
	289	}
	290
	291	error = jhead_scan(jd, &lh_1);
	292	if (error)
	293	return error;
	294
	295	*head = lh_1;
	296
	297	return error;
	298	}
	299
	300	/**
	301	* foreach_descriptor - go through the active part of the log
	302	* @jd: the journal
	303	* @start: the first log header in the active region
	304	* @end: the last log header (don't process the contents of this entry))
	305	*
	306	* Call a given function once for every log descriptor in the active
	307	* portion of the log.
	308	*
	309	* Returns: errno
	310	*/
	311
	312	static int foreach_descriptor(struct gfs2_jdesc *jd, unsigned int start,
	313	unsigned int end, int pass)
	314	{
feaa7bba	315	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	316	struct buffer_head *bh;
	317	struct gfs2_log_descriptor *ld;
	318	int error = 0;
	319	u32 length;
	320	__be64 *ptr;
	321	unsigned int offset = sizeof(struct gfs2_log_descriptor);
a67cdbd4 SW	322	offset += sizeof(__be64) - 1;
a67cdbd4 SW	323	offset &= ~(sizeof(__be64) - 1);
b3b94faa DT	324
	325	while (start != end) {
	326	error = gfs2_replay_read_block(jd, start, &bh);
	327	if (error)
	328	return error;
	329	if (gfs2_meta_check(sdp, bh)) {
	330	brelse(bh);
	331	return -EIO;
	332	}
	333	ld = (struct gfs2_log_descriptor *)bh->b_data;
	334	length = be32_to_cpu(ld->ld_length);
	335
e3167ded	336	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
55167622	337	struct gfs2_log_header_host lh;
b3b94faa DT	338	error = get_log_header(jd, start, &lh);
	339	if (!error) {
	340	gfs2_replay_incr_blk(sdp, &start);
88721877	341	brelse(bh);
b3b94faa DT	342	continue;
	343	}
	344	if (error == 1) {
feaa7bba	345	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	346	error = -EIO;
	347	}
	348	brelse(bh);
	349	return error;
	350	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	351	brelse(bh);
	352	return -EIO;
	353	}
	354	ptr = (__be64 *)(bh->b_data + offset);
	355	error = lops_scan_elements(jd, start, ld, ptr, pass);
	356	if (error) {
	357	brelse(bh);
	358	return error;
	359	}
	360
	361	while (length--)
	362	gfs2_replay_incr_blk(sdp, &start);
	363
	364	brelse(bh);
	365	}
	366
	367	return 0;
	368	}
	369
	370	/**
	371	* clean_journal - mark a dirty journal as being clean
	372	* @sdp: the filesystem
	373	* @jd: the journal
	374	* @gl: the journal's glock
	375	* @head: the head journal to start from
	376	*
	377	* Returns: errno
	378	*/
	379
55167622	380	static int clean_journal(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
b3b94faa	381	{
feaa7bba SW	382	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	383	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	384	unsigned int lblock;
b3b94faa	385	struct gfs2_log_header *lh;
cd915493	386	u32 hash;
b3b94faa DT	387	struct buffer_head *bh;
b3b94faa DT	388	int error;
23591256	389	struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 };
fd88de56	390
b3b94faa DT	391	lblock = head->lh_blkno;
b3b94faa DT	392	gfs2_replay_incr_blk(sdp, &lblock);
23591256	393	bh_map.b_size = 1 << ip->i_inode.i_blkbits;
e9e1ef2b	394	error = gfs2_block_map(&ip->i_inode, lblock, &bh_map, 0);
b3b94faa DT	395	if (error)
b3b94faa DT	396	return error;
7a6bbacb	397	if (!bh_map.b_blocknr) {
b3b94faa DT	398	gfs2_consist_inode(ip);
	399	return -EIO;
	400	}
	401
7a6bbacb	402	bh = sb_getblk(sdp->sd_vfs, bh_map.b_blocknr);
b3b94faa DT	403	lock_buffer(bh);
	404	memset(bh->b_data, 0, bh->b_size);
	405	set_buffer_uptodate(bh);
	406	clear_buffer_dirty(bh);
	407	unlock_buffer(bh);
	408
	409	lh = (struct gfs2_log_header *)bh->b_data;
	410	memset(lh, 0, sizeof(struct gfs2_log_header));
	411	lh->lh_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
88721877	412	lh->lh_header.mh_type = cpu_to_be32(GFS2_METATYPE_LH);
e3167ded	413	lh->lh_header.mh_format = cpu_to_be32(GFS2_FORMAT_LH);
b3b94faa DT	414	lh->lh_sequence = cpu_to_be64(head->lh_sequence + 1);
	415	lh->lh_flags = cpu_to_be32(GFS2_LOG_HEAD_UNMOUNT);
	416	lh->lh_blkno = cpu_to_be32(lblock);
	417	hash = gfs2_disk_hash((const char *)lh, sizeof(struct gfs2_log_header));
	418	lh->lh_hash = cpu_to_be32(hash);
	419
	420	set_buffer_dirty(bh);
	421	if (sync_dirty_buffer(bh))
	422	gfs2_io_error_bh(sdp, bh);
	423	brelse(bh);
	424
	425	return error;
	426	}
	427
da755fdb	428
f057f6cd SW	429	static void gfs2_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
f057f6cd SW	430	unsigned int message)
da755fdb	431	{
f057f6cd SW	432	char env_jid[20];
	433	char env_status[20];
	434	char *envp[] = { env_jid, env_status, NULL };
	435	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
	436	ls->ls_recover_jid_done = jid;
	437	ls->ls_recover_jid_status = message;
	438	sprintf(env_jid, "JID=%d", jid);
	439	sprintf(env_status, "RECOVERY=%s",
	440	message == LM_RD_SUCCESS ? "Done" : "Failed");
	441	kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb SW	442	}
da755fdb SW	443
fe64d517 SW	444	static int gfs2_recover_get_ref(struct slow_work *work)
	445	{
	446	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	447	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
	448	return -EBUSY;
	449	return 0;
	450	}
b3b94faa	451
fe64d517 SW	452	static void gfs2_recover_put_ref(struct slow_work *work)
	453	{
	454	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	455	clear_bit(JDF_RECOVERY, &jd->jd_flags);
	456	smp_mb__after_clear_bit();
	457	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
	458	}
	459
	460	static void gfs2_recover_work(struct slow_work *work)
b3b94faa	461	{
fe64d517	462	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
feaa7bba SW	463	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	464	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
55167622	465	struct gfs2_log_header_host head;
b3b94faa DT	466	struct gfs2_holder j_gh, ji_gh, t_gh;
	467	unsigned long t;
	468	int ro = 0;
	469	unsigned int pass;
	470	int error;
	471
5965b1f4 SW	472	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	473	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
	474	jd->jd_jid);
b3b94faa	475
c78bad11	476	/* Acquire the journal lock so we can do recovery */
b3b94faa	477
5965b1f4 SW	478	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	479	LM_ST_EXCLUSIVE,
	480	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	481	&j_gh);
	482	switch (error) {
	483	case 0:
	484	break;
907b9bce	485
5965b1f4 SW	486	case GLR_TRYFAILED:
	487	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	488	error = 0;
907b9bce	489
5965b1f4 SW	490	default:
	491	goto fail;
	492	};
b3b94faa	493
5965b1f4	494	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8	495	LM_FLAG_NOEXP \| GL_NOCACHE, &ji_gh);
5965b1f4 SW	496	if (error)
	497	goto fail_gunlock_j;
	498	} else {
	499	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	500	}
b3b94faa DT	501
	502	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	503
	504	error = gfs2_jdesc_check(jd);
	505	if (error)
	506	goto fail_gunlock_ji;
	507
	508	error = gfs2_find_jhead(jd, &head);
	509	if (error)
	510	goto fail_gunlock_ji;
	511
	512	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
	513	fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
	514	jd->jd_jid);
	515
	516	t = jiffies;
	517
	518	/* Acquire a shared hold on the transaction lock */
	519
59a1cc6b	520	error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED,
5965b1f4	521	LM_FLAG_NOEXP \| LM_FLAG_PRIORITY \|
6802e340	522	GL_NOCACHE, &t_gh);
b3b94faa DT	523	if (error)
	524	goto fail_gunlock_ji;
	525
	526	if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
	527	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	528	ro = 1;
	529	} else {
7bc5c414 AD	530	if (sdp->sd_vfs->s_flags & MS_RDONLY) {
	531	/* check if device itself is read-only */
	532	ro = bdev_read_only(sdp->sd_vfs->s_bdev);
	533	if (!ro) {
	534	fs_info(sdp, "recovery required on "
	535	"read-only filesystem.\n");
	536	fs_info(sdp, "write access will be "
	537	"enabled during recovery.\n");
	538	}
	539	}
b3b94faa DT	540	}
	541
	542	if (ro) {
7bc5c414 AD	543	fs_warn(sdp, "jid=%u: Can't replay: read-only block "
7bc5c414 AD	544	"device\n", jd->jd_jid);
b3b94faa DT	545	error = -EROFS;
	546	goto fail_gunlock_tr;
	547	}
	548
	549	fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);
	550
	551	for (pass = 0; pass < 2; pass++) {
	552	lops_before_scan(jd, &head, pass);
	553	error = foreach_descriptor(jd, head.lh_tail,
	554	head.lh_blkno, pass);
	555	lops_after_scan(jd, error, pass);
	556	if (error)
	557	goto fail_gunlock_tr;
	558	}
	559
	560	error = clean_journal(jd, &head);
	561	if (error)
	562	goto fail_gunlock_tr;
	563
	564	gfs2_glock_dq_uninit(&t_gh);
5c676f6d	565	t = DIV_ROUND_UP(jiffies - t, HZ);
b3b94faa DT	566	fs_info(sdp, "jid=%u: Journal replayed in %lus\n",
	567	jd->jd_jid, t);
	568	}
	569
5965b1f4 SW	570	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	571	gfs2_glock_dq_uninit(&ji_gh);
b3b94faa	572
f057f6cd	573	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
b3b94faa	574
5965b1f4 SW	575	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid)
5965b1f4 SW	576	gfs2_glock_dq_uninit(&j_gh);
b3b94faa DT	577
b3b94faa DT	578	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
fe64d517	579	return;
b3b94faa	580
5965b1f4	581	fail_gunlock_tr:
b3b94faa	582	gfs2_glock_dq_uninit(&t_gh);
5965b1f4 SW	583	fail_gunlock_ji:
	584	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
	585	gfs2_glock_dq_uninit(&ji_gh);
	586	fail_gunlock_j:
	587	gfs2_glock_dq_uninit(&j_gh);
	588	}
b3b94faa DT	589
	590	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
	591
5965b1f4	592	fail:
f057f6cd	593	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
b3b94faa DT	594	}
b3b94faa DT	595
fe64d517	596	struct slow_work_ops gfs2_recover_ops = {
3d7a641e	597	.owner = THIS_MODULE,
fe64d517 SW	598	.get_ref = gfs2_recover_get_ref,
	599	.put_ref = gfs2_recover_put_ref,
	600	.execute = gfs2_recover_work,
	601	};
2bfb6449	602
2bfb6449	603
fe64d517	604	static int gfs2_recovery_wait(void *word)
b3b94faa	605	{
fe64d517 SW	606	schedule();
fe64d517 SW	607	return 0;
b3b94faa DT	608	}
b3b94faa DT	609
fe64d517	610	int gfs2_recover_journal(struct gfs2_jdesc *jd)
9ac1b4d9	611	{
fe64d517 SW	612	int rv;
	613	rv = slow_work_enqueue(&jd->jd_work);
	614	if (rv)
	615	return rv;
	616	wait_on_bit(&jd->jd_flags, JDF_RECOVERY, gfs2_recovery_wait, TASK_UNINTERRUPTIBLE);
9ac1b4d9 SW	617	return 0;
	618	}
	619