[mirror_ubuntu-jammy-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;
	u64 dblock;
	u32 extlen;
	int error;

	extlen = 32;
	error = gfs2_get_extent(&ip->i_inode, blk, &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);

	head->lh_local_total = be64_to_cpu(lh->lh_local_total);
	head->lh_local_free = be64_to_cpu(lh->lh_local_free);
	head->lh_local_dinodes = be64_to_cpu(lh->lh_local_dinodes);

	return 0;
}
/**
 * get_log_header - read the log header for a given segment
 * @jd: the journal
 * @blk: the block to look at
 * @head: 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))
 * @pass: iteration number (foreach_descriptor() is called in a for() loop)
 *
 * 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);
}

/**
 * update_statfs_inode - Update the master statfs inode or zero out the local
 *			 statfs inode for a given journal.
 * @jd: The journal
 * @head: If NULL, @inode is the local statfs inode and we need to zero it out.
 *	  Otherwise, it @head contains the statfs change info that needs to be
 *	  synced to the master statfs inode (pointed to by @inode).
 * @inode: statfs inode to update.
 */
static int update_statfs_inode(struct gfs2_jdesc *jd,
			       struct gfs2_log_header_host *head,
			       struct inode *inode)
{
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	struct gfs2_inode *ip;
	struct buffer_head *bh;
	struct gfs2_statfs_change_host sc;
	int error = 0;

	BUG_ON(!inode);
	ip = GFS2_I(inode);

	error = gfs2_meta_inode_buffer(ip, &bh);
	if (error)
		goto out;

	spin_lock(&sdp->sd_statfs_spin);

	if (head) { /* Update the master statfs inode */
		gfs2_statfs_change_in(&sc, bh->b_data + sizeof(struct gfs2_dinode));
		sc.sc_total += head->lh_local_total;
		sc.sc_free += head->lh_local_free;
		sc.sc_dinodes += head->lh_local_dinodes;
		gfs2_statfs_change_out(&sc, bh->b_data + sizeof(struct gfs2_dinode));

		fs_info(sdp, "jid=%u: Updated master statfs Total:%lld, "
			"Free:%lld, Dinodes:%lld after change "
			"[%+lld,%+lld,%+lld]\n", jd->jd_jid, sc.sc_total,
			sc.sc_free, sc.sc_dinodes, head->lh_local_total,
			head->lh_local_free, head->lh_local_dinodes);
	} else { /* Zero out the local statfs inode */
		memset(bh->b_data + sizeof(struct gfs2_dinode), 0,
		       sizeof(struct gfs2_statfs_change));
		/* If it's our own journal, reset any in-memory changes too */
		if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) {
			memset(&sdp->sd_statfs_local, 0,
			       sizeof(struct gfs2_statfs_change_host));
		}
	}
	spin_unlock(&sdp->sd_statfs_spin);

	mark_buffer_dirty(bh);
	brelse(bh);
	gfs2_inode_metasync(ip->i_gl);

out:
	return error;
}

/**
 * recover_local_statfs - Update the master and local statfs changes for this
 *			  journal.
 *
 * Previously, statfs updates would be read in from the local statfs inode and
 * synced to the master statfs inode during recovery.
 *
 * We now use the statfs updates in the journal head to update the master statfs
 * inode instead of reading in from the local statfs inode. To preserve backward
 * compatibility with kernels that can't do this, we still need to keep the
 * local statfs inode up to date by writing changes to it. At some point in the
 * future, we can do away with the local statfs inodes altogether and keep the
 * statfs changes solely in the journal.
 *
 * @jd: the journal
 * @head: the journal head
 *
 * Returns: errno
 */
static void recover_local_statfs(struct gfs2_jdesc *jd,
				 struct gfs2_log_header_host *head)
{
	int error;
	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);

	if (!head->lh_local_total && !head->lh_local_free
	    && !head->lh_local_dinodes) /* No change */
		goto zero_local;

	 /* First update the master statfs inode with the changes we
	  * found in the journal. */
	error = update_statfs_inode(jd, head, sdp->sd_statfs_inode);
	if (error)
		goto out;

zero_local:
	/* Zero out the local statfs inode so any changes in there
	 * are not re-recovered. */
	error = update_statfs_inode(jd, NULL,
				    find_local_statfs_inode(sdp, jd->jd_jid));
out:
	return;
}

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;
			goto fail;

		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_freeze_lock(sdp, &thaw_gh, LM_FLAG_PRIORITY);
		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 jd_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) {
				up_read(&sdp->sd_log_flush_lock);
				goto fail_gunlock_thaw;
			}
		}

		recover_local_statfs(jd, &head);
		clean_journal(jd, &head);
		up_read(&sdp->sd_log_flush_lock);

		gfs2_freeze_unlock(&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_freeze_unlock(&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;
cd915493 SW	37	u64 dblock;
cd915493 SW	38	u32 extlen;
b3b94faa DT	39	int error;
b3b94faa DT	40
9153dac1 AG	41	extlen = 32;
9153dac1 AG	42	error = gfs2_get_extent(&ip->i_inode, blk, &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
73092698 AD	147	head->lh_local_total = be64_to_cpu(lh->lh_local_total);
	148	head->lh_local_free = be64_to_cpu(lh->lh_local_free);
	149	head->lh_local_dinodes = be64_to_cpu(lh->lh_local_dinodes);
	150
40e0e61e AD	151	return 0;
40e0e61e AD	152	}
b3b94faa DT	153	/**
	154	* get_log_header - read the log header for a given segment
	155	* @jd: the journal
	156	* @blk: the block to look at
c551f66c	157	* @head: the log header to return
b3b94faa DT	158	*
	159	* Read the log header for a given segement in a given journal. Do a few
	160	* sanity checks on it.
	161	*
	162	* Returns: 0 on success,
	163	* 1 if the header was invalid or incomplete,
	164	* errno on error
	165	*/
	166
	167	static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
55167622	168	struct gfs2_log_header_host *head)
b3b94faa	169	{
40e0e61e	170	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	171	struct buffer_head *bh;
b3b94faa DT	172	int error;
	173
	174	error = gfs2_replay_read_block(jd, blk, &bh);
	175	if (error)
	176	return error;
b3b94faa	177
40e0e61e AD	178	error = __get_log_header(sdp, (const struct gfs2_log_header *)bh->b_data,
40e0e61e AD	179	blk, head);
0ff5916a	180	brelse(bh);
b3b94faa	181
0ff5916a	182	return error;
b3b94faa DT	183	}
b3b94faa DT	184
b3b94faa DT	185	/**
	186	* foreach_descriptor - go through the active part of the log
	187	* @jd: the journal
	188	* @start: the first log header in the active region
	189	* @end: the last log header (don't process the contents of this entry))
c551f66c	190	* @pass: iteration number (foreach_descriptor() is called in a for() loop)
b3b94faa DT	191	*
	192	* Call a given function once for every log descriptor in the active
	193	* portion of the log.
	194	*
	195	* Returns: errno
	196	*/
	197
7c70b896	198	static int foreach_descriptor(struct gfs2_jdesc *jd, u32 start,
b3b94faa DT	199	unsigned int end, int pass)
b3b94faa DT	200	{
feaa7bba	201	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	202	struct buffer_head *bh;
	203	struct gfs2_log_descriptor *ld;
	204	int error = 0;
	205	u32 length;
	206	__be64 *ptr;
	207	unsigned int offset = sizeof(struct gfs2_log_descriptor);
a67cdbd4 SW	208	offset += sizeof(__be64) - 1;
a67cdbd4 SW	209	offset &= ~(sizeof(__be64) - 1);
b3b94faa DT	210
	211	while (start != end) {
	212	error = gfs2_replay_read_block(jd, start, &bh);
	213	if (error)
	214	return error;
	215	if (gfs2_meta_check(sdp, bh)) {
	216	brelse(bh);
	217	return -EIO;
	218	}
	219	ld = (struct gfs2_log_descriptor *)bh->b_data;
	220	length = be32_to_cpu(ld->ld_length);
	221
e3167ded	222	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
55167622	223	struct gfs2_log_header_host lh;
b3b94faa DT	224	error = get_log_header(jd, start, &lh);
b3b94faa DT	225	if (!error) {
e1cb6be9	226	gfs2_replay_incr_blk(jd, &start);
88721877	227	brelse(bh);
b3b94faa DT	228	continue;
	229	}
	230	if (error == 1) {
feaa7bba	231	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	232	error = -EIO;
	233	}
	234	brelse(bh);
	235	return error;
	236	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	237	brelse(bh);
	238	return -EIO;
	239	}
	240	ptr = (__be64 *)(bh->b_data + offset);
	241	error = lops_scan_elements(jd, start, ld, ptr, pass);
	242	if (error) {
	243	brelse(bh);
	244	return error;
	245	}
	246
	247	while (length--)
e1cb6be9	248	gfs2_replay_incr_blk(jd, &start);
b3b94faa DT	249
	250	brelse(bh);
	251	}
	252
	253	return 0;
	254	}
	255
	256	/**
	257	* clean_journal - mark a dirty journal as being clean
b3b94faa	258	* @jd: the journal
b3b94faa DT	259	* @head: the head journal to start from
	260	*
	261	* Returns: errno
	262	*/
	263
588bff95 BP	264	static void clean_journal(struct gfs2_jdesc *jd,
588bff95 BP	265	struct gfs2_log_header_host *head)
b3b94faa	266	{
feaa7bba	267	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
7c70b896	268	u32 lblock = head->lh_blkno;
b3b94faa	269
7c70b896	270	gfs2_replay_incr_blk(jd, &lblock);
7c70b896	271	gfs2_write_log_header(sdp, jd, head->lh_sequence + 1, 0, lblock,
c1696fb8 BP	272	GFS2_LOG_HEAD_UNMOUNT \| GFS2_LOG_HEAD_RECOVERY,
c1696fb8 BP	273	REQ_PREFLUSH \| REQ_FUA \| REQ_META \| REQ_SYNC);
19ebc050 AG	274	if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) {
	275	sdp->sd_log_flush_head = lblock;
	276	gfs2_log_incr_head(sdp);
	277	}
b3b94faa DT	278	}
b3b94faa DT	279
da755fdb	280
f057f6cd SW	281	static void gfs2_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
f057f6cd SW	282	unsigned int message)
da755fdb	283	{
f057f6cd SW	284	char env_jid[20];
	285	char env_status[20];
	286	char *envp[] = { env_jid, env_status, NULL };
	287	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
e0c2a9aa	288
f057f6cd SW	289	ls->ls_recover_jid_done = jid;
f057f6cd SW	290	ls->ls_recover_jid_status = message;
3566c964	291	sprintf(env_jid, "JID=%u", jid);
f057f6cd SW	292	sprintf(env_status, "RECOVERY=%s",
	293	message == LM_RD_SUCCESS ? "Done" : "Failed");
	294	kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
e0c2a9aa DT	295
	296	if (sdp->sd_lockstruct.ls_ops->lm_recovery_result)
	297	sdp->sd_lockstruct.ls_ops->lm_recovery_result(sdp, jid, message);
da755fdb SW	298	}
da755fdb SW	299
bedb0f05 AD	300	/**
	301	* update_statfs_inode - Update the master statfs inode or zero out the local
	302	* statfs inode for a given journal.
	303	* @jd: The journal
	304	* @head: If NULL, @inode is the local statfs inode and we need to zero it out.
	305	* Otherwise, it @head contains the statfs change info that needs to be
	306	* synced to the master statfs inode (pointed to by @inode).
	307	* @inode: statfs inode to update.
	308	*/
	309	static int update_statfs_inode(struct gfs2_jdesc *jd,
	310	struct gfs2_log_header_host *head,
	311	struct inode *inode)
	312	{
	313	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	314	struct gfs2_inode *ip;
	315	struct buffer_head *bh;
	316	struct gfs2_statfs_change_host sc;
	317	int error = 0;
	318
	319	BUG_ON(!inode);
	320	ip = GFS2_I(inode);
	321
	322	error = gfs2_meta_inode_buffer(ip, &bh);
	323	if (error)
	324	goto out;
	325
	326	spin_lock(&sdp->sd_statfs_spin);
	327
	328	if (head) { /* Update the master statfs inode */
	329	gfs2_statfs_change_in(&sc, bh->b_data + sizeof(struct gfs2_dinode));
	330	sc.sc_total += head->lh_local_total;
	331	sc.sc_free += head->lh_local_free;
	332	sc.sc_dinodes += head->lh_local_dinodes;
	333	gfs2_statfs_change_out(&sc, bh->b_data + sizeof(struct gfs2_dinode));
	334
	335	fs_info(sdp, "jid=%u: Updated master statfs Total:%lld, "
	336	"Free:%lld, Dinodes:%lld after change "
	337	"[%+lld,%+lld,%+lld]\n", jd->jd_jid, sc.sc_total,
	338	sc.sc_free, sc.sc_dinodes, head->lh_local_total,
	339	head->lh_local_free, head->lh_local_dinodes);
	340	} else { /* Zero out the local statfs inode */
	341	memset(bh->b_data + sizeof(struct gfs2_dinode), 0,
	342	sizeof(struct gfs2_statfs_change));
	343	/* If it's our own journal, reset any in-memory changes too */
	344	if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) {
	345	memset(&sdp->sd_statfs_local, 0,
	346	sizeof(struct gfs2_statfs_change_host));
	347	}
	348	}
	349	spin_unlock(&sdp->sd_statfs_spin);
	350
	351	mark_buffer_dirty(bh);
	352	brelse(bh);
4a55752a	353	gfs2_inode_metasync(ip->i_gl);
bedb0f05 AD	354
	355	out:
	356	return error;
	357	}
	358
	359	/**
	360	* recover_local_statfs - Update the master and local statfs changes for this
	361	* journal.
	362	*
	363	* Previously, statfs updates would be read in from the local statfs inode and
	364	* synced to the master statfs inode during recovery.
	365	*
	366	* We now use the statfs updates in the journal head to update the master statfs
	367	* inode instead of reading in from the local statfs inode. To preserve backward
	368	* compatibility with kernels that can't do this, we still need to keep the
	369	* local statfs inode up to date by writing changes to it. At some point in the
	370	* future, we can do away with the local statfs inodes altogether and keep the
	371	* statfs changes solely in the journal.
	372	*
	373	* @jd: the journal
	374	* @head: the journal head
	375	*
	376	* Returns: errno
	377	*/
	378	static void recover_local_statfs(struct gfs2_jdesc *jd,
	379	struct gfs2_log_header_host *head)
	380	{
	381	int error;
	382	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	383
	384	if (!head->lh_local_total && !head->lh_local_free
	385	&& !head->lh_local_dinodes) /* No change */
	386	goto zero_local;
	387
	388	/* First update the master statfs inode with the changes we
	389	* found in the journal. */
	390	error = update_statfs_inode(jd, head, sdp->sd_statfs_inode);
	391	if (error)
	392	goto out;
	393
	394	zero_local:
	395	/* Zero out the local statfs inode so any changes in there
	396	* are not re-recovered. */
	397	error = update_statfs_inode(jd, NULL,
	398	find_local_statfs_inode(sdp, jd->jd_jid));
	399	out:
	400	return;
	401	}
	402
6ecd7c2d	403	void gfs2_recover_func(struct work_struct *work)
b3b94faa	404	{
fe64d517	405	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
feaa7bba SW	406	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	407	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
55167622	408	struct gfs2_log_header_host head;
24972557	409	struct gfs2_holder j_gh, ji_gh, thaw_gh;
5e86d9d1	410	ktime_t t_start, t_jlck, t_jhd, t_tlck, t_rep;
b3b94faa DT	411	int ro = 0;
b3b94faa DT	412	unsigned int pass;
4a772772	413	int error = 0;
c741c455	414	int jlocked = 0;
b3b94faa	415
03678a99 BP	416	if (gfs2_withdrawn(sdp)) {
	417	fs_err(sdp, "jid=%u: Recovery not attempted due to withdraw.\n",
	418	jd->jd_jid);
	419	goto fail;
	420	}
5e86d9d1	421	t_start = ktime_get();
4a772772 BP	422	if (sdp->sd_args.ar_spectator)
	423	goto fail;
	424	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
5965b1f4 SW	425	fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
5965b1f4 SW	426	jd->jd_jid);
c741c455	427	jlocked = 1;
c78bad11	428	/* Acquire the journal lock so we can do recovery */
b3b94faa	429
5965b1f4 SW	430	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	431	LM_ST_EXCLUSIVE,
	432	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	433	&j_gh);
	434	switch (error) {
	435	case 0:
	436	break;
907b9bce	437
5965b1f4 SW	438	case GLR_TRYFAILED:
	439	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	440	error = 0;
e5966cf2	441	goto fail;
907b9bce	442
5965b1f4 SW	443	default:
5965b1f4 SW	444	goto fail;
098b9c14	445	}
b3b94faa	446
5965b1f4	447	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8	448	LM_FLAG_NOEXP \| GL_NOCACHE, &ji_gh);
5965b1f4 SW	449	if (error)
	450	goto fail_gunlock_j;
	451	} else {
	452	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	453	}
b3b94faa	454
5e86d9d1	455	t_jlck = ktime_get();
b3b94faa DT	456	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	457
	458	error = gfs2_jdesc_check(jd);
	459	if (error)
	460	goto fail_gunlock_ji;
	461
f4686c26	462	error = gfs2_find_jhead(jd, &head, true);
b3b94faa DT	463	if (error)
b3b94faa DT	464	goto fail_gunlock_ji;
5e86d9d1	465	t_jhd = ktime_get();
98583b3e AD	466	fs_info(sdp, "jid=%u: Journal head lookup took %lldms\n", jd->jd_jid,
98583b3e AD	467	ktime_ms_delta(t_jhd, t_jlck));
b3b94faa DT	468
	469	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
	470	fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
	471	jd->jd_jid);
	472
24972557	473	/* Acquire a shared hold on the freeze lock */
b3b94faa	474
c77b52c0	475	error = gfs2_freeze_lock(sdp, &thaw_gh, LM_FLAG_PRIORITY);
b3b94faa DT	476	if (error)
	477	goto fail_gunlock_ji;
	478
e8ca5cc5 DT	479	if (test_bit(SDF_RORECOVERY, &sdp->sd_flags)) {
	480	ro = 1;
	481	} else if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
b3b94faa DT	482	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	483	ro = 1;
	484	} else {
bc98a42c	485	if (sb_rdonly(sdp->sd_vfs)) {
7bc5c414 AD	486	/* check if device itself is read-only */
	487	ro = bdev_read_only(sdp->sd_vfs->s_bdev);
	488	if (!ro) {
	489	fs_info(sdp, "recovery required on "
	490	"read-only filesystem.\n");
	491	fs_info(sdp, "write access will be "
	492	"enabled during recovery.\n");
	493	}
	494	}
b3b94faa DT	495	}
	496
	497	if (ro) {
7bc5c414 AD	498	fs_warn(sdp, "jid=%u: Can't replay: read-only block "
7bc5c414 AD	499	"device\n", jd->jd_jid);
b3b94faa	500	error = -EROFS;
24972557	501	goto fail_gunlock_thaw;
b3b94faa DT	502	}
b3b94faa DT	503
5e86d9d1	504	t_tlck = ktime_get();
49eb776e BP	505	fs_info(sdp, "jid=%u: Replaying journal...0x%x to 0x%x\n",
49eb776e BP	506	jd->jd_jid, head.lh_tail, head.lh_blkno);
b3b94faa	507
c9ebc4b7 BP	508	/* We take the sd_log_flush_lock here primarily to prevent log
c9ebc4b7 BP	509	* flushes and simultaneous journal replays from stomping on
82218943	510	* each other wrt jd_log_bio. */
c953a735	511	down_read(&sdp->sd_log_flush_lock);
b3b94faa DT	512	for (pass = 0; pass < 2; pass++) {
	513	lops_before_scan(jd, &head, pass);
	514	error = foreach_descriptor(jd, head.lh_tail,
	515	head.lh_blkno, pass);
	516	lops_after_scan(jd, error, pass);
834ec3e1 AG	517	if (error) {
834ec3e1 AG	518	up_read(&sdp->sd_log_flush_lock);
24972557	519	goto fail_gunlock_thaw;
834ec3e1	520	}
b3b94faa DT	521	}
b3b94faa DT	522
bedb0f05	523	recover_local_statfs(jd, &head);
588bff95	524	clean_journal(jd, &head);
c953a735	525	up_read(&sdp->sd_log_flush_lock);
b3b94faa	526
c77b52c0	527	gfs2_freeze_unlock(&thaw_gh);
5e86d9d1 AD	528	t_rep = ktime_get();
	529	fs_info(sdp, "jid=%u: Journal replayed in %lldms [jlck:%lldms, "
	530	"jhead:%lldms, tlck:%lldms, replay:%lldms]\n",
	531	jd->jd_jid, ktime_ms_delta(t_rep, t_start),
	532	ktime_ms_delta(t_jlck, t_start),
	533	ktime_ms_delta(t_jhd, t_jlck),
	534	ktime_ms_delta(t_tlck, t_jhd),
	535	ktime_ms_delta(t_rep, t_tlck));
b3b94faa DT	536	}
b3b94faa DT	537
f057f6cd	538	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
b3b94faa	539
c741c455 SW	540	if (jlocked) {
c741c455 SW	541	gfs2_glock_dq_uninit(&ji_gh);
5965b1f4	542	gfs2_glock_dq_uninit(&j_gh);
c741c455	543	}
b3b94faa DT	544
b3b94faa DT	545	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
6ecd7c2d	546	goto done;
b3b94faa	547
24972557	548	fail_gunlock_thaw:
c77b52c0	549	gfs2_freeze_unlock(&thaw_gh);
5965b1f4	550	fail_gunlock_ji:
c741c455	551	if (jlocked) {
5965b1f4 SW	552	gfs2_glock_dq_uninit(&ji_gh);
	553	fail_gunlock_j:
	554	gfs2_glock_dq_uninit(&j_gh);
	555	}
b3b94faa DT	556
b3b94faa DT	557	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
5965b1f4	558	fail:
376d3778	559	jd->jd_recover_error = error;
f057f6cd	560	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
6ecd7c2d TH	561	done:
6ecd7c2d TH	562	clear_bit(JDF_RECOVERY, &jd->jd_flags);
4e857c58	563	smp_mb__after_atomic();
6ecd7c2d	564	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
b3b94faa DT	565	}
b3b94faa DT	566
6ecd7c2d	567	int gfs2_recover_journal(struct gfs2_jdesc *jd, bool wait)
9ac1b4d9	568	{
fe64d517	569	int rv;
6ecd7c2d TH	570
	571	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
	572	return -EBUSY;
	573
	574	/* we have JDF_RECOVERY, queue should always succeed */
	575	rv = queue_work(gfs_recovery_wq, &jd->jd_work);
	576	BUG_ON(!rv);
	577
	578	if (wait)
74316201	579	wait_on_bit(&jd->jd_flags, JDF_RECOVERY,
6ecd7c2d TH	580	TASK_UNINTERRUPTIBLE);
6ecd7c2d TH	581
376d3778	582	return wait ? jd->jd_recover_error : 0;
9ac1b4d9 SW	583	}
9ac1b4d9 SW	584