[linux-block.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 *gfs2_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 = NULL, *iter;

	list_for_each_entry(iter, head, rr_list) {
		if (iter->rr_blkno == blkno) {
			rr = iter;
			break;
		}
	}

	if (rr) {
		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 = NULL, *iter;
	int wrap, a, b, revoke;

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

	if (!rr)
		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)
{
	const u32 zero = 0;
	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(hash, &zero, 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);

	gfs2_replay_incr_blk(jd, &head->lh_blkno);
	head->lh_sequence++;
	gfs2_write_log_header(sdp, jd, head->lh_sequence, 0, head->lh_blkno,
			      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);
}

/**
 * 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;
	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_withdrawing_or_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 glock...\n",
			jd->jd_jid);
		jlocked = 1;
		/* Acquire the journal glock 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);
	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)) {
		mutex_lock(&sdp->sd_freeze_mutex);

		if (test_bit(SDF_FROZEN, &sdp->sd_flags)) {
			mutex_unlock(&sdp->sd_freeze_mutex);
			fs_warn(sdp, "jid=%u: Can't replay: filesystem "
				"is frozen\n", jd->jd_jid);
			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_nofreeze;
		}

		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_nofreeze;
			}
		}

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

		mutex_unlock(&sdp->sd_freeze_mutex);
		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));
	}

	if (jd->jd_jid == sdp->sd_lockstruct.ls_jid)
		gfs2_log_pointers_init(sdp, &head);

	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_nofreeze:
	mutex_unlock(&sdp->sd_freeze_mutex);
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(gfs2_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;
}

void gfs2_log_pointers_init(struct gfs2_sbd *sdp,
			    struct gfs2_log_header_host *head)
{
	sdp->sd_log_sequence = head->lh_sequence + 1;
	gfs2_replay_incr_blk(sdp->sd_jdesc, &head->lh_blkno);
	sdp->sd_log_tail = head->lh_blkno;
	sdp->sd_log_flush_head = head->lh_blkno;
	sdp->sd_log_flush_tail = head->lh_blkno;
	sdp->sd_log_head = head->lh_blkno;
}
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
5c0dc371	30	struct workqueue_struct *gfs2_recovery_wq;
6ecd7c2d	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;
b846f2d7	58	struct gfs2_revoke_replay rr = NULL, iter;
b3b94faa	59
b846f2d7 JK	60	list_for_each_entry(iter, head, rr_list) {
	61	if (iter->rr_blkno == blkno) {
	62	rr = iter;
b3b94faa DT	63	break;
	64	}
	65	}
	66
b846f2d7	67	if (rr) {
b3b94faa DT	68	rr->rr_where = where;
	69	return 0;
	70	}
	71
16c5f06f	72	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
b3b94faa DT	73	if (!rr)
	74	return -ENOMEM;
	75
	76	rr->rr_blkno = blkno;
	77	rr->rr_where = where;
	78	list_add(&rr->rr_list, head);
	79
	80	return 1;
	81	}
	82
a17d758b	83	int gfs2_revoke_check(struct gfs2_jdesc *jd, u64 blkno, unsigned int where)
b3b94faa	84	{
b846f2d7	85	struct gfs2_revoke_replay rr = NULL, iter;
b3b94faa	86	int wrap, a, b, revoke;
b3b94faa	87
b846f2d7 JK	88	list_for_each_entry(iter, &jd->jd_revoke_list, rr_list) {
	89	if (iter->rr_blkno == blkno) {
	90	rr = iter;
b3b94faa DT	91	break;
	92	}
	93	}
	94
b846f2d7	95	if (!rr)
b3b94faa DT	96	return 0;
b3b94faa DT	97
a17d758b BP	98	wrap = (rr->rr_where < jd->jd_replay_tail);
a17d758b BP	99	a = (jd->jd_replay_tail < where);
b3b94faa DT	100	b = (where < rr->rr_where);
	101	revoke = (wrap) ? (a \|\| b) : (a && b);
	102
	103	return revoke;
	104	}
	105
a17d758b	106	void gfs2_revoke_clean(struct gfs2_jdesc *jd)
b3b94faa	107	{
a17d758b	108	struct list_head *head = &jd->jd_revoke_list;
b3b94faa DT	109	struct gfs2_revoke_replay *rr;
	110
	111	while (!list_empty(head)) {
969183bc	112	rr = list_first_entry(head, struct gfs2_revoke_replay, rr_list);
b3b94faa DT	113	list_del(&rr->rr_list);
	114	kfree(rr);
	115	}
	116	}
	117
40e0e61e AD	118	int __get_log_header(struct gfs2_sbd sdp, const struct gfs2_log_header lh,
	119	unsigned int blkno, struct gfs2_log_header_host *head)
	120	{
b6ccde39	121	const u32 zero = 0;
40e0e61e AD	122	u32 hash, crc;
	123
	124	if (lh->lh_header.mh_magic != cpu_to_be32(GFS2_MAGIC) \|\|
	125	lh->lh_header.mh_type != cpu_to_be32(GFS2_METATYPE_LH) \|\|
	126	(blkno && be32_to_cpu(lh->lh_blkno) != blkno))
	127	return 1;
	128
	129	hash = crc32(~0, lh, LH_V1_SIZE - 4);
b6ccde39	130	hash = ~crc32(hash, &zero, 4); /* assume lh_hash is zero */
40e0e61e AD	131
	132	if (be32_to_cpu(lh->lh_hash) != hash)
	133	return 1;
	134
	135	crc = crc32c(~0, (void *)lh + LH_V1_SIZE + 4,
	136	sdp->sd_sb.sb_bsize - LH_V1_SIZE - 4);
	137
	138	if ((lh->lh_crc != 0 && be32_to_cpu(lh->lh_crc) != crc))
	139	return 1;
	140
	141	head->lh_sequence = be64_to_cpu(lh->lh_sequence);
	142	head->lh_flags = be32_to_cpu(lh->lh_flags);
	143	head->lh_tail = be32_to_cpu(lh->lh_tail);
	144	head->lh_blkno = be32_to_cpu(lh->lh_blkno);
	145
73092698 AD	146	head->lh_local_total = be64_to_cpu(lh->lh_local_total);
	147	head->lh_local_free = be64_to_cpu(lh->lh_local_free);
	148	head->lh_local_dinodes = be64_to_cpu(lh->lh_local_dinodes);
	149
40e0e61e AD	150	return 0;
40e0e61e AD	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
c551f66c	156	* @head: the log header to return
b3b94faa DT	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
b3b94faa DT	184	/**
	185	* foreach_descriptor - go through the active part of the log
	186	* @jd: the journal
	187	* @start: the first log header in the active region
	188	* @end: the last log header (don't process the contents of this entry))
c551f66c	189	* @pass: iteration number (foreach_descriptor() is called in a for() loop)
b3b94faa DT	190	*
	191	* Call a given function once for every log descriptor in the active
	192	* portion of the log.
	193	*
	194	* Returns: errno
	195	*/
	196
7c70b896	197	static int foreach_descriptor(struct gfs2_jdesc *jd, u32 start,
b3b94faa DT	198	unsigned int end, int pass)
b3b94faa DT	199	{
feaa7bba	200	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa DT	201	struct buffer_head *bh;
	202	struct gfs2_log_descriptor *ld;
	203	int error = 0;
	204	u32 length;
	205	__be64 *ptr;
	206	unsigned int offset = sizeof(struct gfs2_log_descriptor);
a67cdbd4 SW	207	offset += sizeof(__be64) - 1;
a67cdbd4 SW	208	offset &= ~(sizeof(__be64) - 1);
b3b94faa DT	209
	210	while (start != end) {
	211	error = gfs2_replay_read_block(jd, start, &bh);
	212	if (error)
	213	return error;
	214	if (gfs2_meta_check(sdp, bh)) {
	215	brelse(bh);
	216	return -EIO;
	217	}
	218	ld = (struct gfs2_log_descriptor *)bh->b_data;
	219	length = be32_to_cpu(ld->ld_length);
	220
e3167ded	221	if (be32_to_cpu(ld->ld_header.mh_type) == GFS2_METATYPE_LH) {
55167622	222	struct gfs2_log_header_host lh;
b3b94faa DT	223	error = get_log_header(jd, start, &lh);
b3b94faa DT	224	if (!error) {
e1cb6be9	225	gfs2_replay_incr_blk(jd, &start);
88721877	226	brelse(bh);
b3b94faa DT	227	continue;
	228	}
	229	if (error == 1) {
feaa7bba	230	gfs2_consist_inode(GFS2_I(jd->jd_inode));
b3b94faa DT	231	error = -EIO;
	232	}
	233	brelse(bh);
	234	return error;
	235	} else if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LD)) {
	236	brelse(bh);
	237	return -EIO;
	238	}
	239	ptr = (__be64 *)(bh->b_data + offset);
	240	error = lops_scan_elements(jd, start, ld, ptr, pass);
	241	if (error) {
	242	brelse(bh);
	243	return error;
	244	}
	245
	246	while (length--)
e1cb6be9	247	gfs2_replay_incr_blk(jd, &start);
b3b94faa DT	248
	249	brelse(bh);
	250	}
	251
	252	return 0;
	253	}
	254
	255	/**
	256	* clean_journal - mark a dirty journal as being clean
b3b94faa	257	* @jd: the journal
b3b94faa DT	258	* @head: the head journal to start from
	259	*
	260	* Returns: errno
	261	*/
	262
588bff95 BP	263	static void clean_journal(struct gfs2_jdesc *jd,
588bff95 BP	264	struct gfs2_log_header_host *head)
b3b94faa	265	{
feaa7bba	266	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
b3b94faa	267
2ebb94ab AG	268	gfs2_replay_incr_blk(jd, &head->lh_blkno);
	269	head->lh_sequence++;
	270	gfs2_write_log_header(sdp, jd, head->lh_sequence, 0, head->lh_blkno,
c1696fb8 BP	271	GFS2_LOG_HEAD_UNMOUNT \| GFS2_LOG_HEAD_RECOVERY,
c1696fb8 BP	272	REQ_PREFLUSH \| REQ_FUA \| REQ_META \| REQ_SYNC);
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
bedb0f05 AD	295	/**
	296	* update_statfs_inode - Update the master statfs inode or zero out the local
	297	* statfs inode for a given journal.
	298	* @jd: The journal
	299	* @head: If NULL, @inode is the local statfs inode and we need to zero it out.
	300	* Otherwise, it @head contains the statfs change info that needs to be
	301	* synced to the master statfs inode (pointed to by @inode).
	302	* @inode: statfs inode to update.
	303	*/
	304	static int update_statfs_inode(struct gfs2_jdesc *jd,
	305	struct gfs2_log_header_host *head,
	306	struct inode *inode)
	307	{
	308	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	309	struct gfs2_inode *ip;
	310	struct buffer_head *bh;
	311	struct gfs2_statfs_change_host sc;
	312	int error = 0;
	313
	314	BUG_ON(!inode);
	315	ip = GFS2_I(inode);
	316
	317	error = gfs2_meta_inode_buffer(ip, &bh);
	318	if (error)
	319	goto out;
	320
	321	spin_lock(&sdp->sd_statfs_spin);
	322
	323	if (head) { /* Update the master statfs inode */
	324	gfs2_statfs_change_in(&sc, bh->b_data + sizeof(struct gfs2_dinode));
	325	sc.sc_total += head->lh_local_total;
	326	sc.sc_free += head->lh_local_free;
	327	sc.sc_dinodes += head->lh_local_dinodes;
	328	gfs2_statfs_change_out(&sc, bh->b_data + sizeof(struct gfs2_dinode));
	329
	330	fs_info(sdp, "jid=%u: Updated master statfs Total:%lld, "
	331	"Free:%lld, Dinodes:%lld after change "
	332	"[%+lld,%+lld,%+lld]\n", jd->jd_jid, sc.sc_total,
	333	sc.sc_free, sc.sc_dinodes, head->lh_local_total,
	334	head->lh_local_free, head->lh_local_dinodes);
	335	} else { /* Zero out the local statfs inode */
	336	memset(bh->b_data + sizeof(struct gfs2_dinode), 0,
	337	sizeof(struct gfs2_statfs_change));
	338	/* If it's our own journal, reset any in-memory changes too */
	339	if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) {
	340	memset(&sdp->sd_statfs_local, 0,
	341	sizeof(struct gfs2_statfs_change_host));
	342	}
	343	}
	344	spin_unlock(&sdp->sd_statfs_spin);
	345
	346	mark_buffer_dirty(bh);
	347	brelse(bh);
4a55752a	348	gfs2_inode_metasync(ip->i_gl);
bedb0f05 AD	349
	350	out:
	351	return error;
	352	}
	353
	354	/**
	355	* recover_local_statfs - Update the master and local statfs changes for this
	356	* journal.
	357	*
	358	* Previously, statfs updates would be read in from the local statfs inode and
	359	* synced to the master statfs inode during recovery.
	360	*
	361	* We now use the statfs updates in the journal head to update the master statfs
	362	* inode instead of reading in from the local statfs inode. To preserve backward
	363	* compatibility with kernels that can't do this, we still need to keep the
	364	* local statfs inode up to date by writing changes to it. At some point in the
	365	* future, we can do away with the local statfs inodes altogether and keep the
	366	* statfs changes solely in the journal.
	367	*
	368	* @jd: the journal
	369	* @head: the journal head
	370	*
	371	* Returns: errno
	372	*/
	373	static void recover_local_statfs(struct gfs2_jdesc *jd,
	374	struct gfs2_log_header_host *head)
	375	{
	376	int error;
	377	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
	378
	379	if (!head->lh_local_total && !head->lh_local_free
	380	&& !head->lh_local_dinodes) /* No change */
	381	goto zero_local;
	382
	383	/* First update the master statfs inode with the changes we
	384	* found in the journal. */
	385	error = update_statfs_inode(jd, head, sdp->sd_statfs_inode);
	386	if (error)
	387	goto out;
	388
	389	zero_local:
	390	/* Zero out the local statfs inode so any changes in there
	391	* are not re-recovered. */
	392	error = update_statfs_inode(jd, NULL,
	393	find_local_statfs_inode(sdp, jd->jd_jid));
	394	out:
	395	return;
	396	}
	397
6ecd7c2d	398	void gfs2_recover_func(struct work_struct *work)
b3b94faa	399	{
fe64d517	400	struct gfs2_jdesc *jd = container_of(work, struct gfs2_jdesc, jd_work);
feaa7bba SW	401	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
feaa7bba SW	402	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
55167622	403	struct gfs2_log_header_host head;
b77b4a48	404	struct gfs2_holder j_gh, ji_gh;
5e86d9d1	405	ktime_t t_start, t_jlck, t_jhd, t_tlck, t_rep;
b3b94faa DT	406	int ro = 0;
b3b94faa DT	407	unsigned int pass;
4a772772	408	int error = 0;
c741c455	409	int jlocked = 0;
b3b94faa	410
4d927b03	411	if (gfs2_withdrawing_or_withdrawn(sdp)) {
03678a99 BP	412	fs_err(sdp, "jid=%u: Recovery not attempted due to withdraw.\n",
	413	jd->jd_jid);
	414	goto fail;
	415	}
5e86d9d1	416	t_start = ktime_get();
4a772772 BP	417	if (sdp->sd_args.ar_spectator)
	418	goto fail;
	419	if (jd->jd_jid != sdp->sd_lockstruct.ls_jid) {
af1abe11	420	fs_info(sdp, "jid=%u: Trying to acquire journal glock...\n",
5965b1f4	421	jd->jd_jid);
c741c455	422	jlocked = 1;
af1abe11	423	/* Acquire the journal glock so we can do recovery */
b3b94faa	424
5965b1f4 SW	425	error = gfs2_glock_nq_num(sdp, jd->jd_jid, &gfs2_journal_glops,
	426	LM_ST_EXCLUSIVE,
	427	LM_FLAG_NOEXP \| LM_FLAG_TRY \| GL_NOCACHE,
	428	&j_gh);
	429	switch (error) {
	430	case 0:
	431	break;
907b9bce	432
5965b1f4 SW	433	case GLR_TRYFAILED:
	434	fs_info(sdp, "jid=%u: Busy\n", jd->jd_jid);
	435	error = 0;
e5966cf2	436	goto fail;
907b9bce	437
5965b1f4 SW	438	default:
5965b1f4 SW	439	goto fail;
098b9c14	440	}
b3b94faa	441
5965b1f4	442	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8	443	LM_FLAG_NOEXP \| GL_NOCACHE, &ji_gh);
5965b1f4 SW	444	if (error)
	445	goto fail_gunlock_j;
	446	} else {
	447	fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
	448	}
b3b94faa	449
5e86d9d1	450	t_jlck = ktime_get();
b3b94faa DT	451	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
	452
	453	error = gfs2_jdesc_check(jd);
	454	if (error)
	455	goto fail_gunlock_ji;
	456
e320050e	457	error = gfs2_find_jhead(jd, &head);
b3b94faa DT	458	if (error)
b3b94faa DT	459	goto fail_gunlock_ji;
5e86d9d1	460	t_jhd = ktime_get();
98583b3e AD	461	fs_info(sdp, "jid=%u: Journal head lookup took %lldms\n", jd->jd_jid,
98583b3e AD	462	ktime_ms_delta(t_jhd, t_jlck));
b3b94faa DT	463
b3b94faa DT	464	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
b77b4a48	465	mutex_lock(&sdp->sd_freeze_mutex);
b3b94faa	466
5432af15	467	if (test_bit(SDF_FROZEN, &sdp->sd_flags)) {
b77b4a48 AG	468	mutex_unlock(&sdp->sd_freeze_mutex);
	469	fs_warn(sdp, "jid=%u: Can't replay: filesystem "
	470	"is frozen\n", jd->jd_jid);
b3b94faa	471	goto fail_gunlock_ji;
b77b4a48	472	}
b3b94faa	473
e8ca5cc5 DT	474	if (test_bit(SDF_RORECOVERY, &sdp->sd_flags)) {
	475	ro = 1;
	476	} else if (test_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags)) {
b3b94faa DT	477	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
	478	ro = 1;
	479	} else {
bc98a42c	480	if (sb_rdonly(sdp->sd_vfs)) {
7bc5c414 AD	481	/* check if device itself is read-only */
	482	ro = bdev_read_only(sdp->sd_vfs->s_bdev);
	483	if (!ro) {
	484	fs_info(sdp, "recovery required on "
	485	"read-only filesystem.\n");
	486	fs_info(sdp, "write access will be "
	487	"enabled during recovery.\n");
	488	}
	489	}
b3b94faa DT	490	}
	491
	492	if (ro) {
7bc5c414 AD	493	fs_warn(sdp, "jid=%u: Can't replay: read-only block "
7bc5c414 AD	494	"device\n", jd->jd_jid);
b3b94faa	495	error = -EROFS;
b77b4a48	496	goto fail_gunlock_nofreeze;
b3b94faa DT	497	}
b3b94faa DT	498
5e86d9d1	499	t_tlck = ktime_get();
49eb776e BP	500	fs_info(sdp, "jid=%u: Replaying journal...0x%x to 0x%x\n",
49eb776e BP	501	jd->jd_jid, head.lh_tail, head.lh_blkno);
b3b94faa	502
c9ebc4b7 BP	503	/* We take the sd_log_flush_lock here primarily to prevent log
c9ebc4b7 BP	504	* flushes and simultaneous journal replays from stomping on
82218943	505	* each other wrt jd_log_bio. */
c953a735	506	down_read(&sdp->sd_log_flush_lock);
b3b94faa DT	507	for (pass = 0; pass < 2; pass++) {
	508	lops_before_scan(jd, &head, pass);
	509	error = foreach_descriptor(jd, head.lh_tail,
	510	head.lh_blkno, pass);
	511	lops_after_scan(jd, error, pass);
834ec3e1 AG	512	if (error) {
834ec3e1 AG	513	up_read(&sdp->sd_log_flush_lock);
b77b4a48	514	goto fail_gunlock_nofreeze;
834ec3e1	515	}
b3b94faa DT	516	}
b3b94faa DT	517
bedb0f05	518	recover_local_statfs(jd, &head);
588bff95	519	clean_journal(jd, &head);
c953a735	520	up_read(&sdp->sd_log_flush_lock);
b3b94faa	521
b77b4a48	522	mutex_unlock(&sdp->sd_freeze_mutex);
5e86d9d1 AD	523	t_rep = ktime_get();
	524	fs_info(sdp, "jid=%u: Journal replayed in %lldms [jlck:%lldms, "
	525	"jhead:%lldms, tlck:%lldms, replay:%lldms]\n",
	526	jd->jd_jid, ktime_ms_delta(t_rep, t_start),
	527	ktime_ms_delta(t_jlck, t_start),
	528	ktime_ms_delta(t_jhd, t_jlck),
	529	ktime_ms_delta(t_tlck, t_jhd),
	530	ktime_ms_delta(t_rep, t_tlck));
b3b94faa DT	531	}
b3b94faa DT	532
93bd5edb AG	533	if (jd->jd_jid == sdp->sd_lockstruct.ls_jid)
	534	gfs2_log_pointers_init(sdp, &head);
	535
f057f6cd	536	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
b3b94faa	537
c741c455 SW	538	if (jlocked) {
c741c455 SW	539	gfs2_glock_dq_uninit(&ji_gh);
5965b1f4	540	gfs2_glock_dq_uninit(&j_gh);
c741c455	541	}
b3b94faa DT	542
b3b94faa DT	543	fs_info(sdp, "jid=%u: Done\n", jd->jd_jid);
6ecd7c2d	544	goto done;
b3b94faa	545
b77b4a48 AG	546	fail_gunlock_nofreeze:
b77b4a48 AG	547	mutex_unlock(&sdp->sd_freeze_mutex);
5965b1f4	548	fail_gunlock_ji:
c741c455	549	if (jlocked) {
5965b1f4 SW	550	gfs2_glock_dq_uninit(&ji_gh);
	551	fail_gunlock_j:
	552	gfs2_glock_dq_uninit(&j_gh);
	553	}
b3b94faa DT	554
b3b94faa DT	555	fs_info(sdp, "jid=%u: %s\n", jd->jd_jid, (error) ? "Failed" : "Done");
5965b1f4	556	fail:
376d3778	557	jd->jd_recover_error = error;
f057f6cd	558	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_GAVEUP);
6ecd7c2d TH	559	done:
6ecd7c2d TH	560	clear_bit(JDF_RECOVERY, &jd->jd_flags);
4e857c58	561	smp_mb__after_atomic();
6ecd7c2d	562	wake_up_bit(&jd->jd_flags, JDF_RECOVERY);
b3b94faa DT	563	}
b3b94faa DT	564
6ecd7c2d	565	int gfs2_recover_journal(struct gfs2_jdesc *jd, bool wait)
9ac1b4d9	566	{
fe64d517	567	int rv;
6ecd7c2d TH	568
	569	if (test_and_set_bit(JDF_RECOVERY, &jd->jd_flags))
	570	return -EBUSY;
	571
	572	/* we have JDF_RECOVERY, queue should always succeed */
5c0dc371	573	rv = queue_work(gfs2_recovery_wq, &jd->jd_work);
6ecd7c2d TH	574	BUG_ON(!rv);
	575
	576	if (wait)
74316201	577	wait_on_bit(&jd->jd_flags, JDF_RECOVERY,
6ecd7c2d TH	578	TASK_UNINTERRUPTIBLE);
6ecd7c2d TH	579
376d3778	580	return wait ? jd->jd_recover_error : 0;
9ac1b4d9 SW	581	}
9ac1b4d9 SW	582
8a43d218 AG	583	void gfs2_log_pointers_init(struct gfs2_sbd *sdp,
8a43d218 AG	584	struct gfs2_log_header_host *head)
703a4af3	585	{
8a43d218 AG	586	sdp->sd_log_sequence = head->lh_sequence + 1;
	587	gfs2_replay_incr_blk(sdp->sd_jdesc, &head->lh_blkno);
	588	sdp->sd_log_tail = head->lh_blkno;
	589	sdp->sd_log_flush_head = head->lh_blkno;
	590	sdp->sd_log_flush_tail = head->lh_blkno;
	591	sdp->sd_log_head = head->lh_blkno;
703a4af3	592	}