[mirror_ubuntu-hirsute-kernel.git] / fs / gfs2 / recovery.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
 * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
 */

#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_first_entry(head, 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;
}

/**
 * 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, u32 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);
	u32 lblock = head->lh_blkno;

	gfs2_replay_incr_blk(jd, &lblock);
	gfs2_write_log_header(sdp, jd, head->lh_sequence + 1, 0, lblock,
			      GFS2_LOG_HEAD_UNMOUNT | GFS2_LOG_HEAD_RECOVERY,
			      REQ_PREFLUSH | REQ_FUA | REQ_META | REQ_SYNC);
	if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) {
		sdp->sd_log_flush_head = lblock;
		gfs2_log_incr_head(sdp);
	}
}


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;

	if (gfs2_withdrawn(sdp)) {
		fs_err(sdp, "jid=%u: Recovery not attempted due to withdraw.\n",
		       jd->jd_jid);
		goto fail;
	}
	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, true);
	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 |
					   GL_EXACT, &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...0x%x to 0x%x\n",
			jd->jd_jid, head.lh_tail, head.lh_blkno);

		/* We take the sd_log_flush_lock here primarily to prevent log
		 * flushes and simultaneous journal replays from stomping on
		 * each other wrt sd_log_bio. */
		down_read(&sdp->sd_log_flush_lock);
		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);
		up_read(&sdp->sd_log_flush_lock);

		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
7336d0e6	1	// SPDX-License-Identifier: GPL-2.0-only
b3b94faa DT	2	/*
b3b94faa DT	3	* Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3a8a9a10	4	* Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
b3b94faa DT	5	*/
b3b94faa DT	6
1c2ea8a2	7	#include <linux/module.h>
b3b94faa DT	8	#include <linux/slab.h>
	9	#include <linux/spinlock.h>
	10	#include <linux/completion.h>
	11	#include <linux/buffer_head.h>
5c676f6d	12	#include <linux/gfs2_ondisk.h>
71b86f56	13	#include <linux/crc32.h>
c1696fb8	14	#include <linux/crc32c.h>
5e86d9d1	15	#include <linux/ktime.h>
b3b94faa DT	16
b3b94faa DT	17	#include "gfs2.h"
5c676f6d	18	#include "incore.h"
b3b94faa DT	19	#include "bmap.h"
	20	#include "glock.h"
	21	#include "glops.h"
588bff95	22	#include "log.h"
b3b94faa DT	23	#include "lops.h"
	24	#include "meta_io.h"
	25	#include "recovery.h"
	26	#include "super.h"
5c676f6d	27	#include "util.h"
71b86f56	28	#include "dir.h"
b3b94faa	29
6ecd7c2d TH	30	struct workqueue_struct *gfs_recovery_wq;
6ecd7c2d TH	31
b3b94faa DT	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	37	int new = 0;
cd915493 SW	38	u64 dblock;
cd915493 SW	39	u32 extlen;
b3b94faa DT	40	int error;
b3b94faa DT	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
7276b3b0	50	*bh = gfs2_meta_ra(gl, dblock, extlen);
b3b94faa DT	51
	52	return error;
	53	}
	54
a17d758b	55	int gfs2_revoke_add(struct gfs2_jdesc *jd, u64 blkno, unsigned int where)
b3b94faa	56	{
a17d758b	57	struct list_head *head = &jd->jd_revoke_list;
b3b94faa DT	58	struct gfs2_revoke_replay *rr;
	59	int found = 0;
	60
	61	list_for_each_entry(rr, head, rr_list) {
	62	if (rr->rr_blkno == blkno) {
	63	found = 1;
	64	break;
	65	}
	66	}
	67
	68	if (found) {
	69	rr->rr_where = where;
	70	return 0;
	71	}
	72
16c5f06f	73	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
b3b94faa DT	74	if (!rr)
	75	return -ENOMEM;
	76
	77	rr->rr_blkno = blkno;
	78	rr->rr_where = where;
	79	list_add(&rr->rr_list, head);
	80
	81	return 1;
	82	}
	83
a17d758b	84	int gfs2_revoke_check(struct gfs2_jdesc *jd, u64 blkno, unsigned int where)
b3b94faa DT	85	{
	86	struct gfs2_revoke_replay *rr;
	87	int wrap, a, b, revoke;
	88	int found = 0;
	89
a17d758b	90	list_for_each_entry(rr, &jd->jd_revoke_list, rr_list) {
b3b94faa DT	91	if (rr->rr_blkno == blkno) {
	92	found = 1;
	93	break;
	94	}
	95	}
	96
	97	if (!found)
	98	return 0;
	99
a17d758b BP	100	wrap = (rr->rr_where < jd->jd_replay_tail);
a17d758b BP	101	a = (jd->jd_replay_tail < where);
b3b94faa DT	102	b = (where < rr->rr_where);
	103	revoke = (wrap) ? (a \|\| b) : (a && b);
	104
	105	return revoke;
	106	}
	107
a17d758b	108	void gfs2_revoke_clean(struct gfs2_jdesc *jd)
b3b94faa	109	{
a17d758b	110	struct list_head *head = &jd->jd_revoke_list;
b3b94faa DT	111	struct gfs2_revoke_replay *rr;
	112
	113	while (!list_empty(head)) {
969183bc	114	rr = list_first_entry(head, struct gfs2_revoke_replay, rr_list);
b3b94faa DT	115	list_del(&rr->rr_list);
	116	kfree(rr);
	117	}
	118	}
	119
40e0e61e AD	120	int __get_log_header(struct gfs2_sbd sdp, const struct gfs2_log_header lh,
	121	unsigned int blkno, struct gfs2_log_header_host *head)
	122	{
	123	u32 hash, crc;
	124
	125	if (lh->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) \|\|
	126	lh->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH) \|\|
	127	(blkno && be32_to_cpu(lh->lh_blkno) != blkno))
	128	return 1;
	129
	130	hash = crc32(~0, lh, LH_V1_SIZE - 4);
	131	hash = ~crc32_le_shift(hash, 4); /* assume lh_hash is zero */
	132
	133	if (be32_to_cpu(lh->lh_hash) != hash)
	134	return 1;
	135
	136	crc = crc32c(~0, (void *)lh + LH_V1_SIZE + 4,
	137	sdp->sd_sb.sb_bsize - LH_V1_SIZE - 4);
	138
	139	if ((lh->lh_crc != 0 && be32_to_cpu(lh->lh_crc) != crc))
	140	return 1;
	141
	142	head->lh_sequence = be64_to_cpu(lh->lh_sequence);
	143	head->lh_flags = be32_to_cpu(lh->lh_flags);
	144	head->lh_tail = be32_to_cpu(lh->lh_tail);
	145	head->lh_blkno = be32_to_cpu(lh->lh_blkno);
	146
	147	return 0;
	148	}
b3b94faa DT	149	/**
	150	* get_log_header - read the log header for a given segment
	151	* @jd: the journal
	152	* @blk: the block to look at
	153	* @lh: the log header to return
	154	*
	155	* Read the log header for a given segement in a given journal. Do a few
	156	* sanity checks on it.
	157	*
	158	* Returns: 0 on success,
	159	* 1 if the header was invalid or incomplete,
	160	* errno on error
	161	*/
	162
	163	static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
55167622	164	struct gfs2_log_header_host *head)
b3b94faa	165	{
40e0e61e	166	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	167	struct buffer_head *bh;
b3b94faa DT	168	int error;
	169
	170	error = gfs2_replay_read_block(jd, blk, &bh);
	171	if (error)
	172	return error;
b3b94faa	173
40e0e61e AD	174	error = __get_log_header(sdp, (const struct gfs2_log_header *)bh->b_data,
40e0e61e AD	175	blk, head);
0ff5916a	176	brelse(bh);
b3b94faa	177
0ff5916a	178	return error;
b3b94faa DT	179	}
b3b94faa DT	180
b3b94faa DT	181	/**
	182	* foreach_descriptor - go through the active part of the log
	183	* @jd: the journal
	184	* @start: the first log header in the active region
	185	* @end: the last log header (don't process the contents of this entry))
	186	*
	187	* Call a given function once for every log descriptor in the active
	188	* portion of the log.
	189	*
	190	* Returns: errno
	191	*/
	192
7c70b896	193	static int foreach_descriptor(struct gfs2_jdesc *jd, u32 start,
b3b94faa DT	194	unsigned int end, int pass)
b3b94faa DT	195	{
feaa7bba	196	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	197	struct buffer_head *bh;
	198	struct gfs2_log_descriptor *ld;
	199	int error = 0;
	200	u32 length;
	201	__be64 *ptr;
	202	unsigned int offset = sizeof(struct gfs2_log_descriptor);
a67cdbd4 SW	203	offset += sizeof(__be64) - 1;
a67cdbd4 SW	204	offset &= ~(sizeof(__be64) - 1);
b3b94faa DT	205
	206	while (start != end) {
	207	error = gfs2_replay_read_block(jd, start, &bh);
	208	if (error)
	209	return error;
	210	if (gfs2_meta_check(sdp, bh)) {
	211	brelse(bh);
	212	return -EIO;
	213	}
	214	ld = (struct gfs2_log_descriptor *)bh->b_data;
	215	length = be32_to_cpu(ld->ld_length);
	216
e3167ded	217	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
55167622	218	struct gfs2_log_header_host lh;
b3b94faa DT	219	error = get_log_header(jd, start, &lh);
b3b94faa DT	220	if (!error) {
e1cb6be9	221	gfs2_replay_incr_blk(jd, &start);
88721877	222	brelse(bh);
b3b94faa DT	223	continue;
	224	}
	225	if (error == 1) {
feaa7bba	226	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	227	error = -EIO;
	228	}
	229	brelse(bh);
	230	return error;
	231	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	232	brelse(bh);
	233	return -EIO;
	234	}
	235	ptr = (__be64 *)(bh->b_data + offset);
	236	error = lops_scan_elements(jd, start, ld, ptr, pass);
	237	if (error) {
	238	brelse(bh);
	239	return error;
	240	}
	241
	242	while (length--)
e1cb6be9	243	gfs2_replay_incr_blk(jd, &start);
b3b94faa DT	244
	245	brelse(bh);
	246	}
	247
	248	return 0;
	249	}
	250
	251	/**
	252	* clean_journal - mark a dirty journal as being clean
b3b94faa	253	* @jd: the journal
b3b94faa DT	254	* @head: the head journal to start from
	255	*
	256	* Returns: errno
	257	*/
	258
588bff95 BP	259	static void clean_journal(struct gfs2_jdesc *jd,
588bff95 BP	260	struct gfs2_log_header_host *head)
b3b94faa	261	{
feaa7bba	262	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
7c70b896	263	u32 lblock = head->lh_blkno;
b3b94faa	264
7c70b896	265	gfs2_replay_incr_blk(jd, &lblock);
7c70b896	266	gfs2_write_log_header(sdp, jd, head->lh_sequence + 1, 0, lblock,
c1696fb8 BP	267	GFS2_LOG_HEAD_UNMOUNT \| GFS2_LOG_HEAD_RECOVERY,
c1696fb8 BP	268	REQ_PREFLUSH \| REQ_FUA \| REQ_META \| REQ_SYNC);
19ebc050 AG	269	if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) {
	270	sdp->sd_log_flush_head = lblock;
	271	gfs2_log_incr_head(sdp);
	272	}
b3b94faa DT	273	}
b3b94faa DT	274
da755fdb	275
f057f6cd SW	276	static void gfs2_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
f057f6cd SW	277	unsigned int message)
da755fdb	278	{
f057f6cd SW	279	char env_jid[20];
	280	char env_status[20];
	281	char *envp[] = { env_jid, env_status, NULL };
	282	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
e0c2a9aa	283
f057f6cd SW	284	ls->ls_recover_jid_done = jid;
f057f6cd SW	285	ls->ls_recover_jid_status = message;
3566c964	286	sprintf(env_jid, "JID=%u", jid);
f057f6cd SW	287	sprintf(env_status, "RECOVERY=%s",
	288	message == LM_RD_SUCCESS ? "Done" : "Failed");
	289	kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
e0c2a9aa DT	290
	291	if (sdp->sd_lockstruct.ls_ops->lm_recovery_result)
	292	sdp->sd_lockstruct.ls_ops->lm_recovery_result(sdp, jid, message);
da755fdb SW	293	}
da755fdb SW	294
6ecd7c2d	295	void gfs2_recover_func(struct work_struct *work)
b3b94faa	296	{
fe64d517	297	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
feaa7bba SW	298	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	299	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
55167622	300	struct gfs2_log_header_host head;
24972557	301	struct gfs2_holder j_gh, ji_gh, thaw_gh;
5e86d9d1	302	ktime_t t_start, t_jlck, t_jhd, t_tlck, t_rep;
b3b94faa DT	303	int ro = 0;
b3b94faa DT	304	unsigned int pass;
4a772772	305	int error = 0;
c741c455	306	int jlocked = 0;
b3b94faa	307
03678a99 BP	308	if (gfs2_withdrawn(sdp)) {
	309	fs_err(sdp, "jid=%u: Recovery not attempted due to withdraw.\n",
	310	jd->jd_jid);
	311	goto fail;
	312	}
5e86d9d1	313	t_start = ktime_get();
4a772772 BP	314	if (sdp->sd_args.ar_spectator)
	315	goto fail;
	316	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
5965b1f4 SW	317	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
5965b1f4 SW	318	jd->jd_jid);
c741c455	319	jlocked = 1;
c78bad11	320	/* Acquire the journal lock so we can do recovery */
b3b94faa	321
5965b1f4 SW	322	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	323	LM_ST_EXCLUSIVE,
	324	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	325	&j_gh);
	326	switch (error) {
	327	case 0:
	328	break;
907b9bce	329
5965b1f4 SW	330	case GLR_TRYFAILED:
	331	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	332	error = 0;
907b9bce	333
5965b1f4 SW	334	default:
5965b1f4 SW	335	goto fail;
098b9c14	336	}
b3b94faa	337
5965b1f4	338	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8	339	LM_FLAG_NOEXP \| GL_NOCACHE, &ji_gh);
5965b1f4 SW	340	if (error)
	341	goto fail_gunlock_j;
	342	} else {
	343	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	344	}
b3b94faa	345
5e86d9d1	346	t_jlck = ktime_get();
b3b94faa DT	347	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	348
	349	error = gfs2_jdesc_check(jd);
	350	if (error)
	351	goto fail_gunlock_ji;
	352
f4686c26	353	error = gfs2_find_jhead(jd, &head, true);
b3b94faa DT	354	if (error)
b3b94faa DT	355	goto fail_gunlock_ji;
5e86d9d1	356	t_jhd = ktime_get();
98583b3e AD	357	fs_info(sdp, "jid=%u: Journal head lookup took %lldms\n", jd->jd_jid,
98583b3e AD	358	ktime_ms_delta(t_jhd, t_jlck));
b3b94faa DT	359
	360	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
	361	fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
	362	jd->jd_jid);
	363
24972557	364	/* Acquire a shared hold on the freeze lock */
b3b94faa	365
24972557	366	error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
623ba664 BP	367	LM_FLAG_NOEXP \| LM_FLAG_PRIORITY \|
623ba664 BP	368	GL_EXACT, &thaw_gh);
b3b94faa DT	369	if (error)
	370	goto fail_gunlock_ji;
	371
e8ca5cc5 DT	372	if (test_bit(SDF_RORECOVERY, &sdp->sd_flags)) {
	373	ro = 1;
	374	} else if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
b3b94faa DT	375	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	376	ro = 1;
	377	} else {
bc98a42c	378	if (sb_rdonly(sdp->sd_vfs)) {
7bc5c414 AD	379	/* check if device itself is read-only */
	380	ro = bdev_read_only(sdp->sd_vfs->s_bdev);
	381	if (!ro) {
	382	fs_info(sdp, "recovery required on "
	383	"read-only filesystem.\n");
	384	fs_info(sdp, "write access will be "
	385	"enabled during recovery.\n");
	386	}
	387	}
b3b94faa DT	388	}
	389
	390	if (ro) {
7bc5c414 AD	391	fs_warn(sdp, "jid=%u: Can't replay: read-only block "
7bc5c414 AD	392	"device\n", jd->jd_jid);
b3b94faa	393	error = -EROFS;
24972557	394	goto fail_gunlock_thaw;
b3b94faa DT	395	}
b3b94faa DT	396
5e86d9d1	397	t_tlck = ktime_get();
49eb776e BP	398	fs_info(sdp, "jid=%u: Replaying journal...0x%x to 0x%x\n",
49eb776e BP	399	jd->jd_jid, head.lh_tail, head.lh_blkno);
b3b94faa	400
c9ebc4b7 BP	401	/* We take the sd_log_flush_lock here primarily to prevent log
	402	* flushes and simultaneous journal replays from stomping on
	403	* each other wrt sd_log_bio. */
c953a735	404	down_read(&sdp->sd_log_flush_lock);
b3b94faa DT	405	for (pass = 0; pass < 2; pass++) {
	406	lops_before_scan(jd, &head, pass);
	407	error = foreach_descriptor(jd, head.lh_tail,
	408	head.lh_blkno, pass);
	409	lops_after_scan(jd, error, pass);
	410	if (error)
24972557	411	goto fail_gunlock_thaw;
b3b94faa DT	412	}
b3b94faa DT	413
588bff95	414	clean_journal(jd, &head);
c953a735	415	up_read(&sdp->sd_log_flush_lock);
b3b94faa	416
24972557	417	gfs2_glock_dq_uninit(&thaw_gh);
5e86d9d1 AD	418	t_rep = ktime_get();
	419	fs_info(sdp, "jid=%u: Journal replayed in %lldms [jlck:%lldms, "
	420	"jhead:%lldms, tlck:%lldms, replay:%lldms]\n",
	421	jd->jd_jid, ktime_ms_delta(t_rep, t_start),
	422	ktime_ms_delta(t_jlck, t_start),
	423	ktime_ms_delta(t_jhd, t_jlck),
	424	ktime_ms_delta(t_tlck, t_jhd),
	425	ktime_ms_delta(t_rep, t_tlck));
b3b94faa DT	426	}
b3b94faa DT	427
f057f6cd	428	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
b3b94faa	429
c741c455 SW	430	if (jlocked) {
c741c455 SW	431	gfs2_glock_dq_uninit(&ji_gh);
5965b1f4	432	gfs2_glock_dq_uninit(&j_gh);
c741c455	433	}
b3b94faa DT	434
b3b94faa DT	435	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
6ecd7c2d	436	goto done;
b3b94faa	437
24972557 BM	438	fail_gunlock_thaw:
24972557 BM	439	gfs2_glock_dq_uninit(&thaw_gh);
5965b1f4	440	fail_gunlock_ji:
c741c455	441	if (jlocked) {
5965b1f4 SW	442	gfs2_glock_dq_uninit(&ji_gh);
	443	fail_gunlock_j:
	444	gfs2_glock_dq_uninit(&j_gh);
	445	}
b3b94faa DT	446
b3b94faa DT	447	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
5965b1f4	448	fail:
376d3778	449	jd->jd_recover_error = error;
f057f6cd	450	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
6ecd7c2d TH	451	done:
6ecd7c2d TH	452	clear_bit(JDF_RECOVERY, &jd->jd_flags);
4e857c58	453	smp_mb__after_atomic();
6ecd7c2d	454	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
b3b94faa DT	455	}
b3b94faa DT	456
6ecd7c2d	457	int gfs2_recover_journal(struct gfs2_jdesc *jd, bool wait)
9ac1b4d9	458	{
fe64d517	459	int rv;
6ecd7c2d TH	460
	461	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
	462	return -EBUSY;
	463
	464	/* we have JDF_RECOVERY, queue should always succeed */
	465	rv = queue_work(gfs_recovery_wq, &jd->jd_work);
	466	BUG_ON(!rv);
	467
	468	if (wait)
74316201	469	wait_on_bit(&jd->jd_flags, JDF_RECOVERY,
6ecd7c2d TH	470	TASK_UNINTERRUPTIBLE);
6ecd7c2d TH	471
376d3778	472	return wait ? jd->jd_recover_error : 0;
9ac1b4d9 SW	473	}
9ac1b4d9 SW	474