journal.h

ocfs1.2.7 源码

/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * journal.h
 *
 * Defines journalling api and structures.
 *
 * Copyright (C) 2003, 2005 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#ifndef OCFS2_JOURNAL_H
#define OCFS2_JOURNAL_H

#include <linux/fs.h>
#include <linux/jbd.h>

enum ocfs2_journal_state {
	OCFS2_JOURNAL_FREE = 0,
	OCFS2_JOURNAL_LOADED,
	OCFS2_JOURNAL_IN_SHUTDOWN,
};

struct ocfs2_super;
struct ocfs2_dinode;
struct ocfs2_journal_handle;

struct ocfs2_journal {
	enum ocfs2_journal_state   j_state;    /* Journals current state   */

	journal_t                 *j_journal; /* The kernels journal type */
	struct inode              *j_inode;   /* Kernel inode pointing to
					       * this journal             */
	struct ocfs2_super        *j_osb;     /* pointer to the super
					       * block for the node
					       * we're currently
					       * running on -- not
					       * necessarily the super
					       * block from the node
					       * which we usually run
					       * from (recovery,
					       * etc)                     */
	struct buffer_head        *j_bh;      /* Journal disk inode block */
	atomic_t                  j_num_trans; /* Number of transactions
					        * currently in the system. */
	unsigned long             j_trans_id;
	struct rw_semaphore       j_trans_barrier;
	wait_queue_head_t         j_checkpointed;

	spinlock_t                j_lock;
	struct list_head          j_la_cleanups;
	struct work_struct        j_recovery_work;
};

extern spinlock_t trans_inc_lock;

/* wrap j_trans_id so we never have it equal to zero. */
static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j)
{
	unsigned long old_id;
	spin_lock(&trans_inc_lock);
	old_id = j->j_trans_id++;
	if (unlikely(!j->j_trans_id))
		j->j_trans_id = 1;
	spin_unlock(&trans_inc_lock);
	return old_id;
}

static inline void ocfs2_set_inode_lock_trans(struct ocfs2_journal *journal,
					      struct inode *inode)
{
	spin_lock(&trans_inc_lock);
	OCFS2_I(inode)->ip_last_trans = journal->j_trans_id;
	spin_unlock(&trans_inc_lock);
}

/* Used to figure out whether it's safe to drop a metadata lock on an
 * inode. Returns true if all the inodes changes have been
 * checkpointed to disk. You should be holding the spinlock on the
 * metadata lock while calling this to be sure that nobody can take
 * the lock and put it on another transaction. */
static inline int ocfs2_inode_fully_checkpointed(struct inode *inode)
{
	int ret;
	struct ocfs2_journal *journal = OCFS2_SB(inode->i_sb)->journal;

	spin_lock(&trans_inc_lock);
	ret = time_after(journal->j_trans_id, OCFS2_I(inode)->ip_last_trans);
	spin_unlock(&trans_inc_lock);
	return ret;
}

/* convenience function to check if an inode is still new (has never
 * hit disk) Will do you a favor and set created_trans = 0 when you've
 * been checkpointed.  returns '1' if the inode is still new. */
static inline int ocfs2_inode_is_new(struct inode *inode)
{
	int ret;

	/* System files are never "new" as they're written out by
	 * mkfs. This helps us early during mount, before we have the
	 * journal open and j_trans_id could be junk. */
	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
		return 0;
	spin_lock(&trans_inc_lock);
	ret = !(time_after(OCFS2_SB(inode->i_sb)->journal->j_trans_id,
			   OCFS2_I(inode)->ip_created_trans));
	if (!ret)
		OCFS2_I(inode)->ip_created_trans = 0;
	spin_unlock(&trans_inc_lock);
	return ret;
}

static inline void ocfs2_inode_set_new(struct ocfs2_super *osb,
				      struct inode *inode)
{
	spin_lock(&trans_inc_lock);
	OCFS2_I(inode)->ip_created_trans = osb->journal->j_trans_id;
	spin_unlock(&trans_inc_lock);
}

extern kmem_cache_t *ocfs2_lock_cache;

struct ocfs2_journal_lock {
	struct inode     *jl_inode;
	struct list_head  jl_lock_list;
};

struct ocfs2_journal_handle {
	handle_t            *k_handle; /* kernel handle.                */
	struct ocfs2_journal        *journal;
	u32                 flags;     /* see flags below.              */
	int                 max_buffs; /* Buffs reserved by this handle */

	/* The following two fields are for ocfs2_handle_add_lock */
	int                 num_locks;
	struct list_head    locks;     /* A bunch of locks to
					* release on commit. This
					* should be a list_head */

	struct list_head     inode_list;
};

#define OCFS2_HANDLE_STARTED			1
/* should we sync-commit this handle? */
#define OCFS2_HANDLE_SYNC			2
static inline int ocfs2_handle_started(struct ocfs2_journal_handle *handle)
{
	return handle->flags & OCFS2_HANDLE_STARTED;
}

static inline void ocfs2_handle_set_sync(struct ocfs2_journal_handle *handle, int sync)
{
	if (sync)
		handle->flags |= OCFS2_HANDLE_SYNC;
	else
		handle->flags &= ~OCFS2_HANDLE_SYNC;
}

/* Exported only for the journal struct init code in super.c. Do not call. */
void ocfs2_complete_recovery(void *data);

/*
 *  Journal Control:
 *  Initialize, Load, Shutdown, Wipe a journal.
 *
 *  ocfs2_journal_init     - Initialize journal structures in the OSB.
 *  ocfs2_journal_load     - Load the given journal off disk. Replay it if
 *                          there's transactions still in there.
 *  ocfs2_journal_shutdown - Shutdown a journal, this will flush all
 *                          uncommitted, uncheckpointed transactions.
 *  ocfs2_journal_wipe     - Wipe transactions from a journal. Optionally
 *                          zero out each block.
 *  ocfs2_recovery_thread  - Perform recovery on a node. osb is our own osb.
 *  ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
 *                          event on.
 *  ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
 */
void   ocfs2_set_journal_params(struct ocfs2_super *osb);
int    ocfs2_journal_init(struct ocfs2_journal *journal,
			  int *dirty);
void   ocfs2_journal_shutdown(struct ocfs2_super *osb);
int    ocfs2_journal_wipe(struct ocfs2_journal *journal,
			  int full);
int    ocfs2_journal_load(struct ocfs2_journal *journal, int local);
int    ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
void   ocfs2_recovery_thread(struct ocfs2_super *osb,
			     int node_num);
int    ocfs2_mark_dead_nodes(struct ocfs2_super *osb);
void   ocfs2_complete_mount_recovery(struct ocfs2_super *osb);

static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb)
{
	atomic_set(&osb->needs_checkpoint, 1);
	wake_up(&osb->checkpoint_event);
}

static inline void ocfs2_checkpoint_inode(struct inode *inode)
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

	if (ocfs2_mount_local(osb))
		return;

	if (!ocfs2_inode_fully_checkpointed(inode)) {
		/* WARNING: This only kicks off a single
		 * checkpoint. If someone races you and adds more
		 * metadata to the journal, you won't know, and will
		 * wind up waiting *alot* longer than necessary. Right
		 * now we only use this in clear_inode so that's
		 * OK. */
		ocfs2_start_checkpoint(osb);

		wait_event(osb->journal->j_checkpointed,