[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/crc32c.h>
#include <linux/ktime.h>

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

struct workqueue_struct *gfs_recovery_wq;

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

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

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

	return error;
}

int gfs2_revoke_add(struct gfs2_jdesc *jd, u64 blkno, unsigned int where)
{
	struct list_head *head = &jd->jd_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_jdesc *jd, u64 blkno, unsigned int where)
{
	struct gfs2_revoke_replay *rr;
	int wrap, a, b, revoke;
	int found = 0;

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

	if (!found)
		return 0;

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

	return revoke;
}

void gfs2_revoke_clean(struct gfs2_jdesc *jd)
{
	struct list_head *head = &jd->jd_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);
	}
}

int __get_log_header(struct gfs2_sbd *sdp, const struct gfs2_log_header *lh,
		     unsigned int blkno, struct gfs2_log_header_host *head)
{
	u32 hash, crc;

	if (lh->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) ||
	    lh->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH) ||
	    (blkno && be32_to_cpu(lh->lh_blkno) != blkno))
		return 1;

	hash = crc32(~0, lh, LH_V1_SIZE - 4);
	hash = ~crc32_le_shift(hash, 4); /* assume lh_hash is zero */

	if (be32_to_cpu(lh->lh_hash) != hash)
		return 1;

	crc = crc32c(~0, (void *)lh + LH_V1_SIZE + 4,
		     sdp->sd_sb.sb_bsize - LH_V1_SIZE - 4);

	if ((lh->lh_crc != 0 && be32_to_cpu(lh->lh_crc) != crc))
		return 1;

	head->lh_sequence = be64_to_cpu(lh->lh_sequence);
	head->lh_flags = be32_to_cpu(lh->lh_flags);
	head->lh_tail = be32_to_cpu(lh->lh_tail);
	head->lh_blkno = be32_to_cpu(lh->lh_blkno);

	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 gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	struct buffer_head *bh;
	int error;

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

	error = __get_log_header(sdp, (const struct gfs2_log_header *)bh->b_data,
				 blk, head);
	brelse(bh);

	return error;
}

/**
 * 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(jd, &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(jd, &start);

		brelse(bh);
	}

	return 0;
}

/**
 * clean_journal - mark a dirty journal as being clean
 * @jd: the journal
 * @head: the head journal to start from
 *
 * Returns: errno
 */

static void clean_journal(struct gfs2_jdesc *jd,
			  struct gfs2_log_header_host *head)
{
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);

	sdp->sd_log_flush_head = head->lh_blkno;
	gfs2_replay_incr_blk(jd, &sdp->sd_log_flush_head);
	gfs2_write_log_header(sdp, jd, head->lh_sequence + 1, 0,
			      GFS2_LOG_HEAD_UNMOUNT | GFS2_LOG_HEAD_RECOVERY,
			      REQ_PREFLUSH | REQ_FUA | REQ_META | REQ_SYNC);
}


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=%u", jid);
	sprintf(env_status, "RECOVERY=%s",
		message == LM_RD_SUCCESS ? "Done" : "Failed");
        kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);

	if (sdp->sd_lockstruct.ls_ops->lm_recovery_result)
		sdp->sd_lockstruct.ls_ops->lm_recovery_result(sdp, jid, message);
}

void gfs2_recover_func(struct work_struct *work)
{
	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	struct gfs2_log_header_host head;
	struct gfs2_holder j_gh, ji_gh, thaw_gh;
	ktime_t t_start, t_jlck, t_jhd, t_tlck, t_rep;
	int ro = 0;
	unsigned int pass;
	int error = 0;
	int jlocked = 0;

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

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

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

		default:
			goto fail;
		};

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

	t_jlck = ktime_get();
	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;
	t_jhd = ktime_get();
	fs_info(sdp, "jid=%u: Journal head lookup took %lldms\n", jd->jd_jid,
		ktime_ms_delta(t_jhd, t_jlck));

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

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

		error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
					   LM_FLAG_NOEXP | LM_FLAG_PRIORITY,
					   &thaw_gh);
		if (error)
			goto fail_gunlock_ji;

		if (test_bit(SDF_RORECOVERY, &sdp->sd_flags)) {
			ro = 1;
		} else if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
			if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
				ro = 1;
		} else {
			if (sb_rdonly(sdp->sd_vfs)) {
				/* 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_thaw;
		}

		t_tlck = ktime_get();
		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_thaw;
		}

		clean_journal(jd, &head);

		gfs2_glock_dq_uninit(&thaw_gh);
		t_rep = ktime_get();
		fs_info(sdp, "jid=%u: Journal replayed in %lldms [jlck:%lldms, "
			"jhead:%lldms, tlck:%lldms, replay:%lldms]\n",
			jd->jd_jid, ktime_ms_delta(t_rep, t_start),
			ktime_ms_delta(t_jlck, t_start),
			ktime_ms_delta(t_jhd, t_jlck),
			ktime_ms_delta(t_tlck, t_jhd),
			ktime_ms_delta(t_rep, t_tlck));
	}

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

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

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

fail_gunlock_thaw:
	gfs2_glock_dq_uninit(&thaw_gh);
fail_gunlock_ji:
	if (jlocked) {
		gfs2_glock_dq_uninit(&ji_gh);
fail_gunlock_j:
		gfs2_glock_dq_uninit(&j_gh);
	}

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

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

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

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

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

	return wait ? jd->jd_recover_error : 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>
c1696fb8	17	#include <linux/crc32c.h>
5e86d9d1	18	#include <linux/ktime.h>
b3b94faa DT	19
b3b94faa DT	20	#include "gfs2.h"
5c676f6d	21	#include "incore.h"
b3b94faa DT	22	#include "bmap.h"
	23	#include "glock.h"
	24	#include "glops.h"
588bff95	25	#include "log.h"
b3b94faa DT	26	#include "lops.h"
	27	#include "meta_io.h"
	28	#include "recovery.h"
	29	#include "super.h"
5c676f6d	30	#include "util.h"
71b86f56	31	#include "dir.h"
b3b94faa	32
6ecd7c2d TH	33	struct workqueue_struct *gfs_recovery_wq;
6ecd7c2d TH	34
b3b94faa DT	35	int gfs2_replay_read_block(struct gfs2_jdesc *jd, unsigned int blk,
	36	struct buffer_head **bh)
	37	{
feaa7bba	38	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
5c676f6d	39	struct gfs2_glock *gl = ip->i_gl;
b3b94faa	40	int new = 0;
cd915493 SW	41	u64 dblock;
cd915493 SW	42	u32 extlen;
b3b94faa DT	43	int error;
b3b94faa DT	44
feaa7bba	45	error = gfs2_extent_map(&ip->i_inode, blk, &new, &dblock, &extlen);
b3b94faa DT	46	if (error)
	47	return error;
	48	if (!dblock) {
5c676f6d	49	gfs2_consist_inode(ip);
b3b94faa DT	50	return -EIO;
	51	}
	52
7276b3b0	53	*bh = gfs2_meta_ra(gl, dblock, extlen);
b3b94faa DT	54
	55	return error;
	56	}
	57
a17d758b	58	int gfs2_revoke_add(struct gfs2_jdesc *jd, u64 blkno, unsigned int where)
b3b94faa	59	{
a17d758b	60	struct list_head *head = &jd->jd_revoke_list;
b3b94faa DT	61	struct gfs2_revoke_replay *rr;
	62	int found = 0;
	63
	64	list_for_each_entry(rr, head, rr_list) {
	65	if (rr->rr_blkno == blkno) {
	66	found = 1;
	67	break;
	68	}
	69	}
	70
	71	if (found) {
	72	rr->rr_where = where;
	73	return 0;
	74	}
	75
16c5f06f	76	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
b3b94faa DT	77	if (!rr)
	78	return -ENOMEM;
	79
	80	rr->rr_blkno = blkno;
	81	rr->rr_where = where;
	82	list_add(&rr->rr_list, head);
	83
	84	return 1;
	85	}
	86
a17d758b	87	int gfs2_revoke_check(struct gfs2_jdesc *jd, u64 blkno, unsigned int where)
b3b94faa DT	88	{
	89	struct gfs2_revoke_replay *rr;
	90	int wrap, a, b, revoke;
	91	int found = 0;
	92
a17d758b	93	list_for_each_entry(rr, &jd->jd_revoke_list, rr_list) {
b3b94faa DT	94	if (rr->rr_blkno == blkno) {
	95	found = 1;
	96	break;
	97	}
	98	}
	99
	100	if (!found)
	101	return 0;
	102
a17d758b BP	103	wrap = (rr->rr_where < jd->jd_replay_tail);
a17d758b BP	104	a = (jd->jd_replay_tail < where);
b3b94faa DT	105	b = (where < rr->rr_where);
	106	revoke = (wrap) ? (a \|\| b) : (a && b);
	107
	108	return revoke;
	109	}
	110
a17d758b	111	void gfs2_revoke_clean(struct gfs2_jdesc *jd)
b3b94faa	112	{
a17d758b	113	struct list_head *head = &jd->jd_revoke_list;
b3b94faa DT	114	struct gfs2_revoke_replay *rr;
	115
	116	while (!list_empty(head)) {
	117	rr = list_entry(head->next, struct gfs2_revoke_replay, rr_list);
	118	list_del(&rr->rr_list);
	119	kfree(rr);
	120	}
	121	}
	122
40e0e61e AD	123	int __get_log_header(struct gfs2_sbd sdp, const struct gfs2_log_header lh,
	124	unsigned int blkno, struct gfs2_log_header_host *head)
	125	{
	126	u32 hash, crc;
	127
	128	if (lh->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) \|\|
	129	lh->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH) \|\|
	130	(blkno && be32_to_cpu(lh->lh_blkno) != blkno))
	131	return 1;
	132
	133	hash = crc32(~0, lh, LH_V1_SIZE - 4);
	134	hash = ~crc32_le_shift(hash, 4); /* assume lh_hash is zero */
	135
	136	if (be32_to_cpu(lh->lh_hash) != hash)
	137	return 1;
	138
	139	crc = crc32c(~0, (void *)lh + LH_V1_SIZE + 4,
	140	sdp->sd_sb.sb_bsize - LH_V1_SIZE - 4);
	141
	142	if ((lh->lh_crc != 0 && be32_to_cpu(lh->lh_crc) != crc))
	143	return 1;
	144
	145	head->lh_sequence = be64_to_cpu(lh->lh_sequence);
	146	head->lh_flags = be32_to_cpu(lh->lh_flags);
	147	head->lh_tail = be32_to_cpu(lh->lh_tail);
	148	head->lh_blkno = be32_to_cpu(lh->lh_blkno);
	149
	150	return 0;
	151	}
b3b94faa DT	152	/**
	153	* get_log_header - read the log header for a given segment
	154	* @jd: the journal
	155	* @blk: the block to look at
	156	* @lh: the log header to return
	157	*
	158	* Read the log header for a given segement in a given journal. Do a few
	159	* sanity checks on it.
	160	*
	161	* Returns: 0 on success,
	162	* 1 if the header was invalid or incomplete,
	163	* errno on error
	164	*/
	165
	166	static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
55167622	167	struct gfs2_log_header_host *head)
b3b94faa	168	{
40e0e61e	169	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	170	struct buffer_head *bh;
b3b94faa DT	171	int error;
	172
	173	error = gfs2_replay_read_block(jd, blk, &bh);
	174	if (error)
	175	return error;
b3b94faa	176
40e0e61e AD	177	error = __get_log_header(sdp, (const struct gfs2_log_header *)bh->b_data,
40e0e61e AD	178	blk, head);
0ff5916a	179	brelse(bh);
b3b94faa	180
0ff5916a	181	return error;
b3b94faa DT	182	}
b3b94faa DT	183
23e93c9b BP	184	/**
	185	* find_good_lh - find a good log header
	186	* @jd: the journal
	187	* @blk: the segment to start searching from
	188	* @lh: the log header to fill in
	189	* @forward: if true search forward in the log, else search backward
	190	*
	191	* Call get_log_header() to get a log header for a segment, but if the
	192	* segment is bad, either scan forward or backward until we find a good one.
	193	*
	194	* Returns: errno
	195	*/
	196
	197	static int find_good_lh(struct gfs2_jdesc jd, unsigned int blk,
	198	struct gfs2_log_header_host *head)
	199	{
	200	unsigned int orig_blk = *blk;
	201	int error;
	202
	203	for (;;) {
	204	error = get_log_header(jd, *blk, head);
	205	if (error <= 0)
	206	return error;
	207
	208	if (++*blk == jd->jd_blocks)
	209	*blk = 0;
	210
	211	if (*blk == orig_blk) {
	212	gfs2_consist_inode(GFS2_I(jd->jd_inode));
	213	return -EIO;
	214	}
	215	}
	216	}
	217
	218	/**
	219	* jhead_scan - make sure we've found the head of the log
	220	* @jd: the journal
	221	* @head: this is filled in with the log descriptor of the head
	222	*
	223	* At this point, seg and lh should be either the head of the log or just
	224	* before. Scan forward until we find the head.
	225	*
	226	* Returns: errno
	227	*/
	228
	229	static int jhead_scan(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
	230	{
	231	unsigned int blk = head->lh_blkno;
	232	struct gfs2_log_header_host lh;
	233	int error;
	234
	235	for (;;) {
	236	if (++blk == jd->jd_blocks)
	237	blk = 0;
	238
	239	error = get_log_header(jd, blk, &lh);
	240	if (error < 0)
	241	return error;
	242	if (error == 1)
	243	continue;
	244
	245	if (lh.lh_sequence == head->lh_sequence) {
	246	gfs2_consist_inode(GFS2_I(jd->jd_inode));
	247	return -EIO;
248	}
249	if (lh.lh_sequence < head->lh_sequence)
250	break;
251
252	*head = lh;
253	}
254
255	return 0;
256	}
257
258	/**
259	* gfs2_find_jhead - find the head of a log
260	* @jd: the journal
261	* @head: the log descriptor for the head of the log is returned here
262	*
263	* Do a binary search of a journal and find the valid log entry with the
264	* highest sequence number. (i.e. the log head)
265	*
266	* Returns: errno
267	*/
268
269	int gfs2_find_jhead(struct gfs2_jdesc jd, struct gfs2_log_header_host head)
270	{
271	struct gfs2_log_header_host lh_1, lh_m;
272	u32 blk_1, blk_2, blk_m;
273	int error;
274
275	blk_1 = 0;
276	blk_2 = jd->jd_blocks - 1;
277
278	for (;;) {
279	blk_m = (blk_1 + blk_2) / 2;
280
281	error = find_good_lh(jd, &blk_1, &lh_1);
282	if (error)
283	return error;
284
285	error = find_good_lh(jd, &blk_m, &lh_m);
286	if (error)
287	return error;
288
289	if (blk_1 == blk_m \|\| blk_m == blk_2)
290	break;
291
292	if (lh_1.lh_sequence <= lh_m.lh_sequence)
293	blk_1 = blk_m;
294	else
295	blk_2 = blk_m;
296	}
297
298	error = jhead_scan(jd, &lh_1);
299	if (error)
300	return error;
301
302	*head = lh_1;
303
304	return error;
305	}
306
b3b94faa DT	307	/**
	308	* foreach_descriptor - go through the active part of the log
	309	* @jd: the journal
	310	* @start: the first log header in the active region
	311	* @end: the last log header (don't process the contents of this entry))
	312	*
	313	* Call a given function once for every log descriptor in the active
	314	* portion of the log.
	315	*
	316	* Returns: errno
	317	*/
	318
	319	static int foreach_descriptor(struct gfs2_jdesc *jd, unsigned int start,
	320	unsigned int end, int pass)
	321	{
feaa7bba	322	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	323	struct buffer_head *bh;
	324	struct gfs2_log_descriptor *ld;
	325	int error = 0;
	326	u32 length;
	327	__be64 *ptr;
	328	unsigned int offset = sizeof(struct gfs2_log_descriptor);
a67cdbd4 SW	329	offset += sizeof(__be64) - 1;
a67cdbd4 SW	330	offset &= ~(sizeof(__be64) - 1);
b3b94faa DT	331
	332	while (start != end) {
	333	error = gfs2_replay_read_block(jd, start, &bh);
	334	if (error)
	335	return error;
	336	if (gfs2_meta_check(sdp, bh)) {
	337	brelse(bh);
	338	return -EIO;
	339	}
	340	ld = (struct gfs2_log_descriptor *)bh->b_data;
	341	length = be32_to_cpu(ld->ld_length);
	342
e3167ded	343	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
55167622	344	struct gfs2_log_header_host lh;
b3b94faa DT	345	error = get_log_header(jd, start, &lh);
b3b94faa DT	346	if (!error) {
e1cb6be9	347	gfs2_replay_incr_blk(jd, &start);
88721877	348	brelse(bh);
b3b94faa DT	349	continue;
	350	}
	351	if (error == 1) {
feaa7bba	352	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	353	error = -EIO;
	354	}
	355	brelse(bh);
	356	return error;
	357	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	358	brelse(bh);
	359	return -EIO;
	360	}
	361	ptr = (__be64 *)(bh->b_data + offset);
	362	error = lops_scan_elements(jd, start, ld, ptr, pass);
	363	if (error) {
	364	brelse(bh);
	365	return error;
	366	}
	367
	368	while (length--)
e1cb6be9	369	gfs2_replay_incr_blk(jd, &start);
b3b94faa DT	370
	371	brelse(bh);
	372	}
	373
	374	return 0;
	375	}
	376
	377	/**
	378	* clean_journal - mark a dirty journal as being clean
b3b94faa	379	* @jd: the journal
b3b94faa DT	380	* @head: the head journal to start from
	381	*
	382	* Returns: errno
	383	*/
	384
588bff95 BP	385	static void clean_journal(struct gfs2_jdesc *jd,
588bff95 BP	386	struct gfs2_log_header_host *head)
b3b94faa	387	{
feaa7bba	388	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	389
588bff95 BP	390	sdp->sd_log_flush_head = head->lh_blkno;
588bff95 BP	391	gfs2_replay_incr_blk(jd, &sdp->sd_log_flush_head);
c1696fb8 BP	392	gfs2_write_log_header(sdp, jd, head->lh_sequence + 1, 0,
	393	GFS2_LOG_HEAD_UNMOUNT \| GFS2_LOG_HEAD_RECOVERY,
	394	REQ_PREFLUSH \| REQ_FUA \| REQ_META \| REQ_SYNC);
b3b94faa DT	395	}
b3b94faa DT	396
da755fdb	397
f057f6cd SW	398	static void gfs2_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
f057f6cd SW	399	unsigned int message)
da755fdb	400	{
f057f6cd SW	401	char env_jid[20];
	402	char env_status[20];
	403	char *envp[] = { env_jid, env_status, NULL };
	404	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
e0c2a9aa	405
f057f6cd SW	406	ls->ls_recover_jid_done = jid;
f057f6cd SW	407	ls->ls_recover_jid_status = message;
3566c964	408	sprintf(env_jid, "JID=%u", jid);
f057f6cd SW	409	sprintf(env_status, "RECOVERY=%s",
	410	message == LM_RD_SUCCESS ? "Done" : "Failed");
	411	kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
e0c2a9aa DT	412
	413	if (sdp->sd_lockstruct.ls_ops->lm_recovery_result)
	414	sdp->sd_lockstruct.ls_ops->lm_recovery_result(sdp, jid, message);
da755fdb SW	415	}
da755fdb SW	416
6ecd7c2d	417	void gfs2_recover_func(struct work_struct *work)
b3b94faa	418	{
fe64d517	419	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
feaa7bba SW	420	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	421	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
55167622	422	struct gfs2_log_header_host head;
24972557	423	struct gfs2_holder j_gh, ji_gh, thaw_gh;
5e86d9d1	424	ktime_t t_start, t_jlck, t_jhd, t_tlck, t_rep;
b3b94faa DT	425	int ro = 0;
b3b94faa DT	426	unsigned int pass;
4a772772	427	int error = 0;
c741c455	428	int jlocked = 0;
b3b94faa	429
5e86d9d1	430	t_start = ktime_get();
4a772772 BP	431	if (sdp->sd_args.ar_spectator)
	432	goto fail;
	433	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
5965b1f4 SW	434	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
5965b1f4 SW	435	jd->jd_jid);
c741c455	436	jlocked = 1;
c78bad11	437	/* Acquire the journal lock so we can do recovery */
b3b94faa	438
5965b1f4 SW	439	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	440	LM_ST_EXCLUSIVE,
	441	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	442	&j_gh);
	443	switch (error) {
	444	case 0:
	445	break;
907b9bce	446
5965b1f4 SW	447	case GLR_TRYFAILED:
	448	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	449	error = 0;
907b9bce	450
5965b1f4 SW	451	default:
	452	goto fail;
	453	};
b3b94faa	454
5965b1f4	455	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8	456	LM_FLAG_NOEXP \| GL_NOCACHE, &ji_gh);
5965b1f4 SW	457	if (error)
	458	goto fail_gunlock_j;
	459	} else {
	460	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	461	}
b3b94faa	462
5e86d9d1	463	t_jlck = ktime_get();
b3b94faa DT	464	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	465
	466	error = gfs2_jdesc_check(jd);
	467	if (error)
	468	goto fail_gunlock_ji;
	469
	470	error = gfs2_find_jhead(jd, &head);
	471	if (error)
	472	goto fail_gunlock_ji;
5e86d9d1	473	t_jhd = ktime_get();
98583b3e AD	474	fs_info(sdp, "jid=%u: Journal head lookup took %lldms\n", jd->jd_jid,
98583b3e AD	475	ktime_ms_delta(t_jhd, t_jlck));
b3b94faa DT	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
24972557	481	/* Acquire a shared hold on the freeze lock */
b3b94faa	482
24972557 BM	483	error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
	484	LM_FLAG_NOEXP \| LM_FLAG_PRIORITY,
	485	&thaw_gh);
b3b94faa DT	486	if (error)
	487	goto fail_gunlock_ji;
	488
e8ca5cc5 DT	489	if (test_bit(SDF_RORECOVERY, &sdp->sd_flags)) {
	490	ro = 1;
	491	} else if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
b3b94faa DT	492	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	493	ro = 1;
	494	} else {
bc98a42c	495	if (sb_rdonly(sdp->sd_vfs)) {
7bc5c414 AD	496	/* check if device itself is read-only */
	497	ro = bdev_read_only(sdp->sd_vfs->s_bdev);
	498	if (!ro) {
	499	fs_info(sdp, "recovery required on "
	500	"read-only filesystem.\n");
	501	fs_info(sdp, "write access will be "
	502	"enabled during recovery.\n");
	503	}
	504	}
b3b94faa DT	505	}
	506
	507	if (ro) {
7bc5c414 AD	508	fs_warn(sdp, "jid=%u: Can't replay: read-only block "
7bc5c414 AD	509	"device\n", jd->jd_jid);
b3b94faa	510	error = -EROFS;
24972557	511	goto fail_gunlock_thaw;
b3b94faa DT	512	}
b3b94faa DT	513
5e86d9d1	514	t_tlck = ktime_get();
b3b94faa DT	515	fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);
	516
	517	for (pass = 0; pass < 2; pass++) {
	518	lops_before_scan(jd, &head, pass);
	519	error = foreach_descriptor(jd, head.lh_tail,
	520	head.lh_blkno, pass);
	521	lops_after_scan(jd, error, pass);
	522	if (error)
24972557	523	goto fail_gunlock_thaw;
b3b94faa DT	524	}
b3b94faa DT	525
588bff95	526	clean_journal(jd, &head);
b3b94faa	527
24972557	528	gfs2_glock_dq_uninit(&thaw_gh);
5e86d9d1 AD	529	t_rep = ktime_get();
	530	fs_info(sdp, "jid=%u: Journal replayed in %lldms [jlck:%lldms, "
	531	"jhead:%lldms, tlck:%lldms, replay:%lldms]\n",
	532	jd->jd_jid, ktime_ms_delta(t_rep, t_start),
	533	ktime_ms_delta(t_jlck, t_start),
	534	ktime_ms_delta(t_jhd, t_jlck),
	535	ktime_ms_delta(t_tlck, t_jhd),
	536	ktime_ms_delta(t_rep, t_tlck));
b3b94faa DT	537	}
b3b94faa DT	538
f057f6cd	539	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
b3b94faa	540
c741c455 SW	541	if (jlocked) {
c741c455 SW	542	gfs2_glock_dq_uninit(&ji_gh);
5965b1f4	543	gfs2_glock_dq_uninit(&j_gh);
c741c455	544	}
b3b94faa DT	545
b3b94faa DT	546	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
6ecd7c2d	547	goto done;
b3b94faa	548
24972557 BM	549	fail_gunlock_thaw:
24972557 BM	550	gfs2_glock_dq_uninit(&thaw_gh);
5965b1f4	551	fail_gunlock_ji:
c741c455	552	if (jlocked) {
5965b1f4 SW	553	gfs2_glock_dq_uninit(&ji_gh);
	554	fail_gunlock_j:
	555	gfs2_glock_dq_uninit(&j_gh);
	556	}
b3b94faa DT	557
b3b94faa DT	558	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
5965b1f4	559	fail:
376d3778	560	jd->jd_recover_error = error;
f057f6cd	561	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
6ecd7c2d TH	562	done:
6ecd7c2d TH	563	clear_bit(JDF_RECOVERY, &jd->jd_flags);
4e857c58	564	smp_mb__after_atomic();
6ecd7c2d	565	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
b3b94faa DT	566	}
b3b94faa DT	567
6ecd7c2d	568	int gfs2_recover_journal(struct gfs2_jdesc *jd, bool wait)
9ac1b4d9	569	{
fe64d517	570	int rv;
6ecd7c2d TH	571
	572	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
	573	return -EBUSY;
	574
	575	/* we have JDF_RECOVERY, queue should always succeed */
	576	rv = queue_work(gfs_recovery_wq, &jd->jd_work);
	577	BUG_ON(!rv);
	578
	579	if (wait)
74316201	580	wait_on_bit(&jd->jd_flags, JDF_RECOVERY,
6ecd7c2d TH	581	TASK_UNINTERRUPTIBLE);
6ecd7c2d TH	582
376d3778	583	return wait ? jd->jd_recover_error : 0;
9ac1b4d9 SW	584	}
9ac1b4d9 SW	585